namumnn. .. g .3
. .1. :5. . «3.. ‘ ‘ ,
A . 5 1% kg»:
,. mum“ fir”. mm”. “Hwy? «J33 .

.31.. ,5
gain .

.31:- h .
Swag—fin
«Mrfitlg nrfflvbflut.

a a
‘ shubéaag
i. .5. 1:... l.
t .31.}... E ..
.Wt.3w.mfl:§. $393
1.... r. d 39.1
c 5"!!!" l "I t— . Al.
.5 Eggs .{nfiflwh
- .iul.‘l|ah
' “.l"1

1|

i.il|nf.rh$!lltl

.|.»I$mli)lllllt:tp
lliéi»! .

1!. .z..i.!:. .16....-
irtehvnfisz,
9.. if .(.(..50. 995:};
3;.vxix-vitxl:

:01:
:1: run '1‘ . .1531!
1::

1.! Illlueu
o. .1. t3].
l

IESIS

 

2. LIBRARY
we Michi an State
‘ ' ' Un varsity

 

 

 

This is to certify that the
thesis entitled

THE ROLE OF CASSAVA IN SMALLHOLDEFI MAIZE MARKETING
IN ZAMBIA AND MOZAMBIQUE

presented by

HUNTER H. NIELSON

has been accepted towards fulfillment
of the requirements for the

Master of degree in Agricultural, Food, and
Science Resource Economics

   

 

 

  

MSU is an Affirmative Action/Equal Opportunity Employer

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 KzlProlecc8Pres/ClRC/DateDue,indd

 

 

THE ROLE OF CASSAVA IN SMALLHOLDER MAIZE MARKETING
IN ZAMBIA AND MOZAMBIQUE

By

Hunter H. Nielson

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE

Agricultural, Food, and Resource Economics

2009

ABSTRACT

THE ROLE OF CASSAVA IN SMALLHOLDER MAIZE MARKETING
IN ZAMBIA AND MOZAMBIQUE

By

Hunter H. Nielson
This study investigates the relationship between cassava production and maize marketing
for smallholder farming households in Zambia and Mozambique. Three different
econometric models are estimated for two years of data each in Zambia and Mozambique
to explore the relationship between cassava hectares and kilograms of maize sold. It is
hypothesized that households with cassava stocks in addition to maize are better able to
meet their staple food consumption needs, thereby freeing surplus maize for sale.
Findings indicate that cassava hectares are positively related to maize market
participation and level of maize sales for two good rainfall years in Zambia. In
Mozambique cassava hectares have little impact on maize sales in a good rainfall year,
and a slightly greater impact during a year of poor rainfall. These results indicate that
cassava hectares are an important determinant for maize sales in Zambia, and imply that
in Mozambique households may depend upon cassava more during years of poor maize

production.

Cepyright by
HUNTER H. NIELSON
2009

ACKNOWLEDGMENTS

I would like to express sincere thanks to my committee, composed of Dr. Steven
Haggblade, Dr. David Tschirley, Dr. Joseph Messina, and Dr. Robert Myers for their
encouragement and support. This study benefitted greatly from their individual expertise
and I am grateful for their mentorship. Dr. J effry Wooldridge helped me tremendously in
understanding the econometric techniques utilized here, and Dr. David Mather assisted

with his time and clear, specific advice.

There is a long list of graduate students in the department to whom I am indebted. I
would like to recognize Bill Burke for allowing me to benefit fi'om his State code which
greatly assisted in the completion of this paper; Tim Komarek for his eagerness to
question, teach and lend a hand; J oleen Hadrich and Nicky Mason for their continued
help and advice in all arenas since the beginning; Jacob Gilbert-Rickers for the help
jumping double hurdles; Jordan Chamberlain for his 618 expertise and Ana Fernandez

for everything else.

iv

TABLE OF CONTENTS

 

 

 

 

 

 

 

 

 

LIST OF TABLES vii
LIST OF FIGURES viii
1. INTRODUCTION 1
1.1 PROBLEM STATEMENT ........................................................................................................... 1
1.2 OBJECTIVES OF THE STUDY .................................................................................................. 3
1.3 Focus COUNTRIES .................................................................................................................. 4
1.4 ORGANIZATION OF THE STUDY ............................................................................................ 4

2. IMPORTANCE OF MAIZE AND CASSAVA 6
2.1 HISTORICAL PERSPECTIVE .................................................................................................... 6
2.2 IMPORTANCE OF MAIZE AND CASSAVA ........................................................................... 10

2. 2. 1 Zambia ................................................................................................................................ 10
2.2.2 Mozambique ...................................................................................................................... 13

2.3 SPATIAL DIMENSIONS OF PRODUCTION AND CONSUMPTION ..................................... 17

2. 3. 1 Climate and Agro-ecological Zones ........................................................................... I 7
2.3.2 Spatial Distribution of Production ............................................................................. 23

3. MARKET PARTICIPATION 30
3.1 IMPORTANCE OF AND FACTORS AFFECTING MARKET PARTICIPATION ................... 30
3.2 SUMMARY OF MAIZE SALES IN ZAMBIA AND MOZAMBIQUE ..................................... 33
3.2.1 Zambia ................................................................................................................................ 34
3.2.2 Mozambique ...................................................................................................................... 39

3.3 FACTORS THAT IMPEDE MAIZE SALES ............................................................................... 4o

3. 3. 1 Physical factors ................................................................................................................ 40

3. 3.2 Political factors ................................................................................................................ 43

3.4 POTENTIAL CONSUMPTION SHOCK ABSORBERS ............................................................. 48

4. DATA & SAMPLING METHODS 51
4.1 ZAMBIA SUPPLEMENTAL SURVEY TO THE POST HARVEST SURVEY ........................ 51
4.2 MOZAMBIQUE NATIONAL AGRICULTURAL HOUSEHOLD SURVEY (TIA) ............... 52
4.3 GEOSPATIAL DATA ............................................................................................................... 53
5: ANALYTICAL APPROACH AND HYPOTHESIS 56
5.1: HOUSEHOLD DECISION MAKING .............................................................................................. 56
5.2: HYPOTHESIS ................................................................................................................................. 58
5.3: KEY VARIABLES FOR HYPOTHESIS TESTING ......................................................................... 58

. EMPIRICAL ANALYSIS 60
6.1 ESTIMATION METHODS FOR LIMITED DEPENDENT VARIABLE MODELS ................. 60
6.2 METHODS ..................................................................................................... 67
6.2.1 Variables ............................................................................................................................ 67
6.2.2 Estimation techniques ..................................................................................................... 71

7. RESULTS

 

7.1 DETAILED DISCUSSION OF PRIMARY VARIABLES .........................................................
7.1.1 Zambia: 1999/00 (good rainfall year) .......................................................................
7.1.2 Zambia 2002/03 (good rainfall year) ........................................................................
7.1.3 Summary .............................................................................................................................
7.1.4 Mozambique 2001/02 (good rainfall year) ..............................................................
7.1.5 Mozambique 2004/05 (below average rainfall year) ............................................
7.1. 6 Summary .............................................................................................................................

7.2 BRIEF DISCUSSION OF AUXILIARY VARIABLES .............................................................

8. CONCLUSIONS

8.1 FINAL COMMENTS .................................................................................................................

APPENDIX

 

 

REFERENCES

 

vi

73
74
74
75
77
78
80
82
83

85
86

97

LIST OF TABLES

TABLE 1: ZAMBIA NATIONAL LEVEL MAIZE AND CASSAVA PRODUCTION ........................... 10
TABLE 2: MOZAMBIQUE NATIONAL MAIZE AND CASSAVA PRODUCTION ..... '. ...................... 14
TABLE 3: ZAMBIA PRODUCTION OF MAIZE AND CASSAVA BY AEZ ................................... 23
TABLE 4: MOZAMBIQUE MAIZE AND CASSAVA PRODUCTION BY AEZ ............................... 27
TABLE 5: MAIZE MARKETING IN ZAMBIA AND MOZAMBIQUE ............................................ 35
TABLE 6: ZAMBIA SPATIAL DISTRIBUTION OF MAIZE SALES ............................................... 36
TABLE 7: MOZAMBIQUE SPATIAL DISTRIBUTION OF SALES ................................................ 40
TABLE 8: ZAMBIA EXPLANATORY VARIABLES AND MEANS ................................................ 69
TABLE 9: MOZAMBIQUE EXPLANATORY VARIABLES AND MEANS ...................................... 70
TABLE 10: ZAMBIA KEY RESULTS ...................................................................................... 76
TABLE 11: MOZAMBIQUE KEY RESULTS ............................................................................. 81
TABLE 12: ZAMBIA 1999/00 COMPLETE RESULTS .............................................................. 89
TABLE 13: ZAMBIA 2002/03 COMPLETE RESULTS ....................................................... I ....... 91
TABLE 14: MOZAMBIQUE 2001/02 COMPLETE RESULTS ..................................................... 93
TABLE 15: MOZAMBIQUE 2004/05 COMPLETE RESULTS ................................................................ 95

vii

LIST OF FIGURES

FIGURE 1: ZAMBIA NATIONAL LEVEL MAIZE AND CASSAVA CONSUMPTION,

KCAL/PERSON/DAY .......................................................................................... 1 1
FIGURE 2: ZAMBIA MAIZE AND CASSAVA PRODUCTION ('OOOS MT) .................................. 12
FIGURE 3: INFLATION ADJUSTED DISTRICT MEDIAN MAIZE PRICES (ZMK/KILOGRAM) ....... 13
FIGURE 4: MOZAMBIQUE MAIZE AND CASSAVA CONSUMPTION KCAL/PERSON/DAY .......... 15
FIGURE 5: MOZAMBIQUE MAIZE AND CASSAVA PRODUCTION ('OOOS MT) ......................... 16
FIGURE 6: INFLATION-ADJUSTED MAIZE PRODUCER PRICES (USD/MT) ............................. 17
FIGURE 7: ZAMBIA AGRO-ECOLOGICAL ZONES ................................................................. 19
FIGURE 8: CLASSIFICATION OF AEZS BASED ON 3O-YEAR AVERAGE RAINFALL ................. 22
FIGURE 9: SPATIAL DISTRIBUTION OF CASSAVA AND MAIZE HECTARES ............................. 26

FIGURE 10: ZAMBIA AND MOZAMBIQUE, PERCENT OF POPULATION LIVING IN URBAN AREAS

........................................................................................................................ 31
FIGURE 11: DISTRIBUTION OF MAIZE SALES BY QUARTILE ................................................. 38
FIGURE 12: ROAD NETWORK AND POPULATION DENSITY .................................................. 42
FIGURE 13: TRADE FLOWS, SOUTHERN AFRICA ................................................................. 47

viii

1. INTRODUCTION

1.1 Problem Statement

Maize and cassava are the two most widely produced food staples in sub-Saharan Africa.
Millions of smallholder farm households rely on these two crops to meet their
consumption needs. Maize is the region’s primary food staple, supplying approximately
325 kcal per person per day on average, while cassava, the second most important food
staple supplies approximately 270 kcal per person per day (FAOSTAT). Historically,
maize has been more politically strategic than cassava and has received the lion’s share of
public investment in production and marketing. Rarely, if ever, do policy makers
consider the important contribution cassava makes to food security, and very little

research has been done to explore the relationships that exist between the two crops.

Compared to cassava, maize is marketed to a higher degree and has been important since
colonial times for feeding both rural and urban populations. In Zambia, approximately
37% of the total value of household income comes from maize while less than 15% of
income comes from cassava (Zulu, 2007). Boosting maize production and increasing
market participation has long been a strategic goal for governments in the region as a way
to feed a rapidly urbanizing population. Unfortunately, attempts to achieve this goal have
met with only partial success. Maize production in Southern Afiica remains well below
potential despite various government attempts to improve it. Likewise, although

governments have tried to invigorate maize markets in the region with policies intended

to increase supply, only about one quarter of smallholder farming households sell their

surplus in any given year.

A number of studies have attempted to shed light on the determinants of household
marketing behavior in developing countries. In most cases, market participation choices
are studied with methods similar to those used to address technology adoption questions
(Barrett, 2007). Factors frequently identified as influential to food marketing include
farm size, capacity to bear risk, human capital, labor availability, access to credit, land
tenure, and market access (Peder, 2007). Recent empirical attempts to identify and
measure the determinants of market participation in Africa have focused on one or

another of these factors (Goetz, 1992; Boughton, 2007; Barrett, 2007).

In broad terms, smallholder food market participation depends on the availability of
household productive assets, relative commodity prices, adequacy of market
infrastructure, and ability to meet food consumption needs. The last of these points is
arguably the most pressing for smallholder maize farmers. At harvest time, a household
assesses their maize harvest to determine if it is adequate to meet the household’s needs
before deciding to sell. Due to highly variable annual maize production, some
households will not have maize surplus available for sale in all years. Additionally,
uncertain availability of maize in local markets and the large wedge between producer
prices and retail prices may lead households to sell only a small portion of their surplus

while some households refrain fiom selling maize at all.

Improving household access to other staple foods such as cassava may increase maize
sales behavior by increasing household food security. Cassava is a drought-resistant
perennial crop that is resistant to the rainfall shortfalls that often disrupt maize
production. In cases in which the maize harvest is low, cassava can substitute for maize
in household cOnsumption. Households that produce cassava in addition to maize can

meet their staple food needs even if the maize harvest is poor.

1.2 Objectives of the Study

Recognizing the importance of maize and cassava to smallholder farmers in SSA — and
the potential for increased market participation to boost economic growth — this study
incorporates cassava as an important explanatory variable in maize market participation
models. Specifically, the study aims to understand the afi‘ect cassava availability has on a
household’s maize marketing decision. Simply put, does cassava — a food substitute for

maize in household consumption — act as a catalyst for maize sales?

Improving food security and boosting maize marketing is a priority for Zambia and
Mozambique, yet market participation studies that empirically estimate the determinants
of participation are few. To specifically incorporate cassava into an empirical analysis of
household maize marketing behavior is unique to this study, and should shed light on,
and increase interest in, the interaction between cassava and maize in the discussion of

rural farm welfare and marketing behavior.

1.3 Focus Countries

To explore these questions empirically, this study focuses on two countries in Southern
Africa: Zambia and Mozambique. These two countries are interesting for a number of
reasons. First, Zambia and Mozambique share similar agro-ecological environments.
Second, both countries are intermittent surplus maize producers and exporters, and have
high potential to meet demand for maize in the Democratic Republic of Congo (for
Zambia) and Southern Malawi (for Mozambique). Third, contrasting institutional and
policy enviromnents allow for interesting comparison. And finally, in a region of the
world with notoriously scarce agricultural data, Zambia and Mozambique are anomalies.
In both nations, national level household agricultural surveys are available that represent
some of the most accurate and in-depth information available on agricultural households

in the region.

1.4 Organization of the Study

Chapter 2 outlines the importance of maize and cassava in Zambia and Mozambique and
places the discussion in a historical context. National level and Spatially disaggregated
summaries of maize and cassava production are presented, and variation by Agro-
ecological zone is discussed. Chapter 3 provides a brief discussion of the importance of
market participation for economic growth, both generally, and in the context of Zambia
and Mozambique. It includes an overview of the current maize marketing situation in the
two countries and describes factors that enable or inhibit full market participation.
Chapter 4 provides a discussion of the survey and geospatial data used in the study.

Chapter 5 outlines the analytical approach and hypothesis. Chapter 6 describes the

empirical models and methods used in the study. Chapter 7 is a presentation of results.

And Chapter 8 provides conclusions and suggestions for further study.

2. IMPORTANCE OF MAIZE AND CASSAVA

2.1 Historical Perspective

The importance of maize as a politically strategic crop in Southern Afiica is rooted in an
historical context extending back to the beginning of the twentieth century with European
colonization. Like many other countries, Zambia’s current policy environment is closely

tied to its history.

In the early 1900’s, Zambia (then Northern Rhodesia) was controlled by the British South
Africa (BSA) Company, whose principal goal was to exploit the land and people of
Zambia for the benefit of shareholders in England. By the early 1920’s, Zambian mining
was expanding, and the BSA Company began to look for reliable sources of food to
provide for the workers and rapidly growing towns. Because Afiican farming systems
had been so badly disrupted by the large scale migration of males to mining centers, the
BSA Company believed that Afiican farming was incapable of meeting demand. This
belief led the BSA Company to seek agricultural goods from European settler farmers

who had been given exclusive access to the best land.

Through the 1920’s, the Northern Rhodesian mines expanded, and the demand for maize
was largely met through private marketing channels. However, by the 1930’s, the global
depression caused the closure of several mines, and the subsequent loss of demand drove
prices downward. Despite having lost access to the prime agricultural land near the

major north-south corridor, Afiican farmers tripled their sales volume from the 1920’s

through the 1930’s while settler farmers expanded by only 25 percent. The increasing
pressure from African farmers — who could produce at lower cost —- and declining prices
led the colonial administration to establish what would be the beginning of a long history
of government intervention in Zambian maize marketing with the establishment of the

Maize Control Board (MCB) in 1936.

The MCB skewed incentives to benefit European producers at the expense of Afiican
producers, established maize as the dominant staple food in the country, encouraging its
production at the expense of other crops, and set the precedent for state intervention in
agriculture that persists today. From this point forward, the government would accept the
responsibility for creating a stable market environment with maize at its core. State
involvement in agriculture took on a range of schemes that involved Africans to a greater
or lesser extent. What remained constant was a bias toward Europeans, farmers near the
line of rail, and maize over other crops. However, fiom the beginning Of European
involvement in Northern Rhodesia, agriculture has had a low priority with respect to
other sectors. Beginning with the BSA Company, mining and urban activities were

considered paramount, with agriculture serving a subsidiary role (Wood, 1990)

After independence in 1964, maize continued to be the single most important crop for
Zambian farmers. Zulu et a1 (2007) describe how the government entered into a “social
contract” with smallholder African farmers to redress the neglect they received under the
colonial government. About three quarters of the population resided on farms and relied

on farming as their primary livelihood. Therefore, the newly formed Zambian

government used maize production incentives and input subsidies as the primary vehicles
to improve rural welfare. From Independence to the 1980s, maize yields rose
dramatically as a consequence, even in northern areas traditionally considered unsuited

for maize production (Kokwe, 1997).

The maize boom came at a very high and ultimately unsustainable cost. Beginning with
new leadership in 1991, pressured by the intemational community to adopt standard
structural adjustment policies, the government stopped the pan-territorial pricing policy
that was common in previous decades. At the same time, government reduced its maize
purchases. AS a result of these dramatic changes to the long standing status quo of
government assistance in maize production and marketing, widespread crop
diversification took place. Farmers moved away from maize toward other crops that had
been marginalized over the years of the maize boom. From 1990 to 1999, maize area

declined 22% while cassava increased 65% over the same time period (Zulu, 2000)

The results of the government cutbacks were controversial in Zambia. In 2001 the
government returned to pan-territorial pricing and introduced the Food Reserve Agency
(FRA), another form of government buying agent. While the volumes of food purchased
by the FRA were reduced in the new millennium the FRA announced its intention to

purchase as much as 25% of the marketed crop (Zulu, 2007).

In Southern Africa, policy reflects and accentuates the importance of maize in the

national economy. Zambia channeled the majority of producer and consumer subsidies

into maize at the expense of other crops (Tschirley et a1, 2006). To this day, even though
the magnitude of intervention has dropped from the post independence years, price
supports and fertilizer subsidies are in place for Zambia to such a degree that nearly 50%
of the total agricultural budget was devoted to the FRA and fertilizer support in 2005

(Jayne, 2007).

The early history of Mozambique’s relationship to maize is similar to that of Zambia’s,
but has tended toward a more market-oriented approach in recent years. Post
independence government policies were highly controlling and included a marketing
parastatal and pan-territorial producer prices. In the early 1990’s, controls were reduced
and a greater degree of privatization was allowed. Fixed producer prices were abolished
and the marketing parastatal AGRICOM had collapsed. The Instituto de Cereais de
Mocambique (ICM) replaced AGRICOM in 1995 and was endowed with a broad
mandate to provide inputs and market information, to maintain a strategic grain reserve,
and to act as the buyer of last resort. However, unlike Zambia’s F RA, ICM essentially
behaved like a private trader and relied exclusively on private bank financing, not

government funding.

In comparison to Zambia, maize production in. Mozambique has remained relatively free
of government interference once the controls were lifted. However, as discussed in
Chapter 3, the strategic importance of maize for both countries continues to influence

government policy with regard to marketing and trade.

2.2 Importance of Maize and Cassava

This section examines the importance of maize and cassava to smallholder farming

households in terms of production and consumption at the national level.

2.2.1 Zambia

In Zambia, maize is the dominant staple crop and cassava ranks a distant second.
According to household surveys conducted for the harvest years 1999/2000 and
2002/2003, about half as many households produce cassava as compared to those which
produce maize (Table 1); this imbalance can largely be attributed to the country’s long
history of production and marketing subsidies favoring maize. In contrast, cassava has

received limited attention only in recent years.

Table 1: Zambia national level maize and cassava production

 

 

 

ZAMBIA - % of HH Growing Mean Froduction (kgIth

Maize Cassava Maize Cassava
1999100 77% 39% 948 578
2002103 83% 47% 857 553

 

* Mean production of those who grew each crop
Source: Zambia Supplemental Surveys

Corresponding to the large proportion of households growing maize compared to cassava,
consumption patterns in Zambia are heavily weighted toward maize, and have been
essentially stable through time. Figure 1 compares maize and cassava consumption using
kilocalories per capita per day at the national level from 1990 through 2003. On a per-
capita basis, maize accounts for roughly six times the daily calorie intake of cassava.

This is likely a reflection of the distinctly regional nature of cassava consumption

10

preferences, that cassava is very rarely sold outside cassava production regions, and that
households that consume cassava are also likely to consume maize.

Figure 1: Zambia national level maize and cassava consumption, Kcal/person/day

 

 

1400

1200 ,_

 

1000 "

 

 

 

800 — — Maize

 

 

Cassava

 

600

 

 

400

 

 

 

200

 

 

 

0 I I T I
N '1. '5 <0 '\ 9
1909199 .99 8°’ eye” $050199 $033199 e°°1°° 1° '9

 

 

 

Source: FAO

Figure 2 below demonstrates that national production has been highly variable
throughout the 1990’s and into the early 2000’s. Because a successful maize harvest in
Zambia is mostly dependent on adequate precipitation, maize production is highly
correlated with rainfall. Extreme fluctuations in weather in Southern Afiica, therefore,
lead to fluctuations in maize output. Global climate change stands to exacerbate this
situation. Jain (2007) notes that dry spells have been on the rise in Southern Afiica, and
the length of the rainfall season has been decreasing. Since 1990, there have been at least
two major droughts in Southern Afiica (Zulu, 2000). Given this environment, the greater
resilience of cassava production to rainfall fluctuations is an important asset for food

security.

11

When looking at national cassava production, it is informative to consider how cassava is
produced. Cassava is a root crop that is not harvested in the first year of its growth. In
the second and third year (and sometimes longer) farmers dig up the roots according to
their consumption needs. Therefore, unlike maize, which is harvested and recorded at
one time, cassava production takes place over time as households collect small quantities
fiom mature fields according to their needs. This is important for two reasons. First, it
complicates the data collection exercise because households are expected to recall and be
able to accurately measure several small harvests of cassava at the time of survey
enumeration. Second, the total harvest figures are much less a reflection of the status of
potential stocks than a reflection of how much cassava a household chose to harvest in a
given year. Therefore, when analyzing cassava production in figure 2, it is worth
remembering that they do not necessarily reflect the total harvestable amount of cassava,
but rather the amount that was needed.

Figure 2: Zambia maize and cassava production ('000s MT)

2500
— — Maize—r
Cassava

 

 

 

 

 

 

 

 

 

 

 

l

1

2000 - » , , .1

I" f

1000 l

l

500] , .

O I ‘l ‘I TI' 7 I I I 7 Tfi—V 7 I 1‘

‘— (‘0 O) V— 1") L0 I\ O) F LO l\ O) v- C') 1.0 N O) ‘— 1'") IO N

838883883888888838838889
Source: FAO

The relative instability of maize is important for the discussion at hand. While maize is
highly productive, the volatility of production, coupled with high and steady demand
shown in Figure 1 creates food insecurity for rural households. Additionally, highly
volatile maize production creates highly volatile farm-gate prices for households
attempting to sell maize. Figure 3 is the inflation adjusted median of district-level maize
prices from 1991/92 to 2003/04. The volatility of prices is a reflection of the volatility of
production. This volatility is passed through the value chain, translating into unstable

retail maize meal prices for urban consumers.

Figure 3: Inflation adjusted district median maize prices (ZMK/kilogram)

 

 

14

 

12

.0 \/\ /\ /\
V \V/ \\..,

 

 

 

 

 

 

 

 

0 I 1 I T I ‘I I j I I T I
9192 9293 9394 9495 9596 9697 9798 9899 9900 0001 0102 0203 0304

 

 

 

Source: Zambia Post Harvest Surveys

2. 2.2 Mozambique

According to Walker et a1 (2006) a large percentage of the rural population in
Mozambique considers maize (49%) and cassava (40%) their primary food staple.
Additionally, about 98% of those who consider cassava and maize their primary staple

also produce the crop themselves. With such a large percentage of the country’s

l3

population relying on their own production to satisfy their need for staple food calories, it

is clear that both maize and cassava are key determinants of Mozambican food security

(Walker, 2006).

The percentage of farming households that produce cassava in Mozambique is quite
different from Zambia. As Table 2 Shows, roughly 80% of households produce maize
and roughly 70% of households produce cassava. Unlike Zambia, Mozambican
agricultural policy has been less biased toward maize over the years, and tastes for

cassava have developed more firlly than in Zambia where only the north of the country

has a long tradition of cassava production.

Table 2: Mozambique national maize and cassava production

 

 

 

 

nozAM' BIQUE % of HH Growing Mean Production (kglhh)*
Maize Cassava Maize Cassava

2001/02 81% 71% 480 709

2004/05 78% 71 % 281 1435

* Mean production of those who grew each crop

Source: TIA

Table 2 also illustrates the productive potential of cassava in years when maize harvests
are below normal. The 2004/2005 harvest was poor in all parts of Mozambique, a fact
reflected in the low mean maize production figures that year compared to the average
rainfall year of 2001/2002. In contrast, cassava production more than doubled from
2001/2002. It is logical that households in Mozambique increase the amount of cassava

they harvest as a way to make up for reduced maize production.

14

Corresponding to the greater prominence of cassava in the production figures above,
national households consumption of cassava is also higher than maize (Figure 4). This is
the opposite situation from Zambia, in which maize was most prominent. This is likely a
reflection of the spatial distribution of population and crop production in Mozambique.
The north of the country is home to 60% of the total rural population, and produces both
cassava and maize in high quantities; the center and south have lower populations and are

less likely to produce cassava.

Figure 4: Mozambique maize and cassava consumption Kcal/person/day

 

900
800
700
600
500
400
300
200
100

0

.909 .39" '99": .965 .99“ ’99” Kg05° $9 ’90? ’99? (190° moo" (196‘:

 

Cassava
—- -- Maize

 

 

 

 

 

‘b
o
'19

 

 

 

Source: FAO

Maize and cassava production in Mozambique is also markedly different from Zambia.
Overall tonnage of cassava in any given year is much greater than maize, and the gap
seems to be widening. Figure 5 does not indicate whether the disparity is due to poor
maize harvests or whether cassava planting has increased more rapidly over the years

than maize.

15

Figure 5: Mozambique maize and cassava production ('000s MT)

 

 

8,0(1)

 

Cassava /\

7'0““ “F _ _ Maize //
6,00)

5,... N
4.08 A /

3.000 W W

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MH
2,0(11
’I‘
1,000 fx~y
~—’~~v’\-~—§’A‘/\f/
' I r7 T TTI IT IT r11
8888888888888888888 ages
v-Fu-FV-Fs-x-n-V-FPFFV-FPFs-s-N NN

 

 

 

Source: FAO

Although the volatility in maize production seen in Zambia is not as evident in
Mozambique, the variation in prices is considerable (Figure 6). Additionally, seasonal
maize deficits lead households to increase their consumption of other crops, including
cassava, dramatically. In one study, households that relied on maize for 60% of their
caloric intake during the post-harvest season reduced their reliance to less than 40%

during the hungry season (Rose, 1999).

The spatial distribution of production in Mozambique likely has an impact on volatility,
and normally prevents large-scale maize shortfalls fi'om occurring. The next section
shows that the central and northern areas of the country produce the bulk of the maize,
and are situated in mid- to high-rainfall zones. The low-lying north of the country

infrequently suffers from poor rainfall which keeps production steadier from year to year.

16

Figure 6: Inflation-adjusted maize producer prices (U SD/MT)

 

160

 

140 //
1:: I /\\ - /
,, t / v V\\/

,, \V/

 

 

 

 

 

 

 

 

 

 

 

 

4o
20 I I I I I I I I I f I F l I
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Source: FAO

2.3 Spatial Dimensions of Production and Consumption

A great deal of variability in the production of cassava and maize is masked in the
national-level discussion of the previous section. For both Zambia and Mozambique, the
relative importance of maize and cassava differs depending on the region of the country.
This section describes the analytically important concept of Agro-ecological Zones

(AEZS) before disaggregating the countries’ production data into AEZS for discussion.

2.3.1 Climate and Agro-ecological Zones

Zambia

In general, northern Zambia receives the highest average annual rainfall and the southern
region and valley areas receive the least. The north enjoys the longest growing season,
while the central and southern regions experience shorter growing seasons, often

characterized by dry spells that negatively impact maize production. The Zambian

l7

Government has divided the country into four distinct agro-ecological regions defined by
rainfall, temperature, and soil type. Below is a description of each AEZ corresponding to

Figure 7.

AEZ 1 includes the Zambezi and Luangwa Valley areas in the south and east. This hot,
low-lying zone is characterized by less than 800 mm of rainfall per annurn (Jain, 2007).
The distribution of rainfall is erratic, drought is recurrent, and under normal conditions
the growing season lasts fi'om 60 — 90 days. The zone is best suited to the production of
small grains and livestock rearing, and is not particularly well suited to maize or cassava.
Nevertheless, maize is grown by a large proportion of small-scale farm households in this
zone owing to the long-standing reliance on maize as the staple food of Zambia.
Agricultural production is typically lower here than in other areas of the country, and
households are likely to depend on food from outside the area to meet their needs for part

of the year.

18

 

 

 

ES 82 5.5 .986 an e on HE D
28 2253 EE 89 - 8» EN Nm<
3:8 53 SE 89 - 08 How N? E
E8 cow .55 mm»... ”P Nm< D

    

 

 

 

 

Figure 7: Zambia Agro-ecological Zones

.r\\.IllIIr

I
W 8 N2

l\rl..

 

 

 

 

 

Source: Adapted from Zambia Central Statistical Office

19

AEZ 2 covers the center of the country, stretching from the western to the eastern
borders, wrapping around the Luangwa Valley in the East. This zone is characterized by
average rainfall between 800 mm and 1000 mm per year and is subdivided by soil type
into two Zones: AEZ 2a and AEZ 2b. This region is highly productive for both food and
cash crops. Region 2b, confined to the far West and the Zambezi flood plain, is made up
of sandy soils and has high potential for cassava and rice production as well as cattle
rearing. Though the zone is prone to periodic drought, the majority of small-scale
farmers here devote a large portion of their available land to maize. The growing season

in AEZ 2 ranges from 90 — 190 days (Jain, 2007).

AEZ 3 is the largest agro-ecological region of the country. It comprises almost half of
the territory of Zambia, stretching from the northwest to the northeast of the country.
AEZ 3 is the highest rainfall zone in the country with average rainfall ranging from 1000
mm to 1400 mm per year. The length of the growing season ranges from 140 - 200 days,
and although AEZ 3 is known as a highly productive cassava growing area, maize is a
major crop here as well. For the purposes of this study, AEZ 3 is broken into two parts to
reflect difference in household crop production and marketing opportunities. Districts
near highly urbanized mining areas have more ready access to inputs, and higher
effective demand for commodities. This area is classified AEZ 3a. AEZ 3b is
characterized by a highly rural population, poor infi'astructure and remoteness from

markets.

20

Mozambique

In an effort to standardize the AEZS between Zambia and Mozambique, Zambia’s
government-established AEZS are used as a template from which the Mozambique zones

are derived (Figure 8). AS noted earlier, the AEZS in Zambia are largely, but not

exclusively determined by rainfall]. For simplicity, rainfall is used to define the AEZS in
Mozambique. In order to categorize rural farming households in Mozambique by agro-
ecological zones that approximately match the definitions used in Zambia, the physical

boundaries of the Zambian AEZS are overlaid on the Worldclim Afiica—wide 30 year

average rainfall dataset using Geographical Information Systems (GIS) softwarez. By
adjusting the classification of the rainfall gradient layer to approximately match the
boundaries of the Zambian AEZS, Mozambican zones are classified. These zones
correspond to the three broadest Zambian AEZs: Zone 1: < 800 mm; Zone 2: >800

<=1000; Zone 3 > 1000.

1 Other factors such as soil composition were also considered, but incorporating them was
beyond the scope of this thesis. The rainfall-based AEZS used here are an approximation,
but appear to be fairly consistent with known cropping patterns.

2ESRI ArcGIS 9. 2 was used for all GIS analysis In this paper.

21

Figure 8: Classification of AEZS based on 30 year average rainfall

 

 

    

finance...
.M

 

{1 .

 

 

 

is... I-
82 rv 8? I
.8... D

r/

Sosa /

 

 

.r\\ff\tl\\r\

 

22

2.3.2 Spatial Distribution of Production

Zambia

Production patterns of maize and cassava in Zambia vary greatly by Agro-ecological
Zone (Table 3). The proportion of households that produce maize is greater than 80% in
every AEZ except for AEZ 3b, and is greater than 90% in the two regions most often
associated with maize production. It is worth remembering that AEZ 3a and AEZ3b
receive comparable rainfall, but do not share infrastructure and market access. This
might account for the dramatic difference between AEZ 3a, in which 81% of households
produced maize, and AEZ 3b in which less than 40% produce maize despite the location

of the two zones in the high rainfall belt.

Table 3: Zambia production of maize and cassava by AEZ

 

 

 

 

 

 

 

fitment _ % of HH Growing Mean Production (kg/h h)*
1999/00 96 of HH in each AEZ Maize Cassava Maize Cassava
AEZ 1 8% 91% 2% 1,098 8
AEZ 2a 38% 97% 5% 2,034 37
AEZ 2b 9% 80% 62% 408 373
AEZ 3a 15% 81% 39% 941 566
AEZ 3b 30% 39% 86% 262 1,908
% of HH Growing Mean Production (kglh h)*
2002/03 % of HH in each AEZ Maize Cassava Maize Cassava
AEZ 1 9% 93% 2% 799 1
AEZ 2a 38% 97% 7% 1.725 29
AEZ 2b 8% 84% 81 % 398 463
AEZ 3a 15% 86% 54% 987 506
AEZ 3b 30% 52% 92% 373 1,765

 

* Mean production of those who grew each crop
Source: Zambia Supplemental Surveys

The percent of households that produce cassava is much higher in the high rainfall zones

(AEZ 3a and AEZ 3b) and the sandy-soils of the mid-rainfall zone (AEZ 2b). In the

19903, the reduction of price supports and abolition of pan-territorial pricing for maize,

23

coupled with the development and dispersion of early-maturing and pest-resistant cassava
planting material led many households to diversify their crop production away from
maize. Cassava production between 1992 and 1998 increased 71% in Northern Province
(located in AEZ 3b) alone (Zulu, 2000). Particularly in AEZ 3b, farmers began to seek

crops such as cassava that could stabilize their staple food supply (Zulu, 2007).

Mean household maize production for those who produce maize is highest in AEZ l,
AEZ 2a, and AEZ 3a. However, in the two years displayed in Table 3, which are similar
in terms of national annual rainfall, notable production variation was experienced in the
two drier zones compared to AEZ 3a. Both AEZ l and AEZ 2a had significantly lower
maize production in the second year. Whether the decline is due to localized dry-spells
which are more common in the lower rainfall zones, or to the timing of planting or access
to inputs, Table 3 demonstrates that production in the areas responsible for the majority
of the nation’s maize can fluctuate considerably. It also shows that production in zones in
the high-rainfall belt can be dramatically different than the low rainfall zones. AEZ 3a
deviated only slightly between the two years, and moved in the opposite direction to AEZ
l and AEZ 2a. AEZ 3b, the remote high rainfall area, experienced a sizeable increase in

the second year.

Mean household production of cassava is at least twice as high in AEZ 3b as in any other
zone. Cassava thrives in warmer, wetter areas, and grows well in sandy soils.
Populations in AEZ 3b have a longer history with cassava and the majority of rural

households consume cassava as their primary staple food. Most of the production is

24

consumed by the same households that produce it. In contrast, the small amount of
cassava produced in AEZ l and AEZ 2 is likely to be sold as snack foods in towns, at bus

stops, and around public events (Haggblade, 2007).

Another way to compare the relative importance of maize and cassava is to compare the
size of landholdings devoted to each crop. In figure 9, dark areas on the map indicate a
high ratio of cassava area to maize area, and light areas on the map indicate a high maize
to cassava ratio3. It is clear that the northern areas of Zambia are dominated by
household cassava production, while the central, southern, and eastern parts of the
country are dominated by maize production. Note that the area in the center-north of the
country has a lower cassava to maize hectares ratio compared to other areas in the high-
rainfall zone, corresponding to the proximity of urban demand created by the mining

industry there.

 

3 In Zambia, districts are represented to display the relative importance of each crop.

25

 

 

 

if).

 

a... .28... O P. vv =8 one). I
a: 28: On. - 5 852.382 .U, . A.
a... .23: O a - 3 283 .63 I

a: 28: O 2- NC 85579688 I
a: 23: O 2 AV :8 $88 I
«238: uses. 32806: «>330

 

 

 

          

 

Spatial 'stribution of cassava and maize hectares

Figure 9

I

r R

s
r. \\ x

I;
\\

 

\

\I

\III
\I.
\IISI >I>(i III II

I

 

WNW

/

 

 

Zambia Supplemental Surveys & Mozambique TIA

Source:

26

Mozambique

The distribution of Mozambique’s smallholder farming population is unevenly split
between AEZS (Table 4). Approximately 60% of all farming households reside in AEZ
3, and 40% reside in AEZ 2 and AEZ 1 combined. The low percentage of farming
households living in AEZ l is a reflection of the poor agricultural conditions of the area.
With the exception of some rice production, AEZ 1 is not highly productive compared to
the other regions. The urban areas of the South are natural markets for maize coming out
of AEZ 1 but due to the climatic disadvantage, the majority of food comes from other
areas. The central regions mostly dominated by AEZ 2 and AEZ 3 cOntribute much of
the maize grain to the urban areas, while the maize for maize meal is imported directly

from South Africa.

Table 4: Mozambique maize and cassava production by AEZ
% of I-TH Growing Mean Production (kqlhhr

 

 

 

 

01/02 % of HH in each AEZ Maize Cassava Maize Cassava
AEZ 1 15% 88% 52% 449 342
AEZ 2 29% 86% 71 % 525 724
AEZ 3 57% 76% 76% 385 657

 

°/o of MH Growingg MeanT’roducfion (kg/hh)*

 

 

 

04/05 % of HM in each AEZ Maize Cassava Maize Cassava
AEZ 1 11% 91% 49% 157 443
AEZ 2 29% 84% 67% 286 1,237
AEZ 3 60% 77% 81 % 286 1,718
* Mean production of those who grew each crop
Source: TIA

Compared to Zambia, the percentage of households growing maize is similar between
AEZS, and is quite high. Over three-quarters of households in all zones produced maize

in both years considered here. However, as in Zambia, the lowest percentage of

27

households producing maize is the wettest zone, AEZ 3. The population distribution of
Mozambique is important to keep in mind when considering the percentage of
households in each zone that produce maize, because although a high proportion of
households in each of the AEZS produce maize, the total quantities produced differ

greatly as a consequence of population and rainfall distribution.

Northern Mozambique, which falls almost entirely in AEZ 3, is often the lone surplus
maize producer in times of regional stress. During the severe regional drought of 1992,
rainfall and production in Northern Mozambique was only mildly affected. Typically

this region produces surplus maize which is consumed within the region or exported to

the large deficit areas of Malawi4. Compared to 2001/2002 harvest season, which was an
average rainfall and production year, the 2004/2005 year saw significant decreases in

' maize production in all three AEZS. However, although households in AEZ 3 produced
on average three quarters as much in 2004/05 compared to 2001/2002, the reduction was

proportionally much smaller than the other two zones (Table 4).

Overall, the percentage of farming households that produce cassava is higher in
Mozambique than in Zambia. The proportion of households producing cassava in high
rainfall Zone is comparable to the high rainfall zone of Zambia, but in AEZ l and AEZ 2
the proportion is much higher. In even the lowest rainfall zone, roughly half of all
households produced cassava in Mozambique. It is interesting that in the 2004/2005,

when rainfall was below average for all zones, the production of cassava increased in

_k

4 Such was the case during the Southern Afiica Food Crisis in 2002.

28

each zone. This is consistent with the manner in which households harvest cassava,
digging up roots as needed. If maize harvests are compromised by poor rainfall, one
would expect households to rely on cassava as a substitute. The magnitude of the

increase was more than double for AEZ 3 and nearly so for AEZ 2.

The production figures in Table 4 illustrate the volatility of maize and the potential for
cassava to fill the gap in times of poor maize production. Maize harvests were decimated
in 2004/2005 in all three zones, and cassava production spiked in response. The
perennial nature of cassava’s maturation process makes it an invaluable food security
crop. Once the plant is established, it is very resilient to dry spells, and is able to be

harvested even when maize performs poorly.

Referring again to the map of cassava and maize area planted above (Figure 9), it is clear

that cassava and maize are grown in conjunction in more areas than in Zambias. Where
in Zambia the north has very high ratios of cassava to maize hectares, and the south has
very high ratios of maize to cassava hectares, the picture in Mozambique is somewhat
less extreme. There are larger areas of “maize mixed” and “dual staple” zones in
Mozambique owing to the fact that so many households in all regions of the country
produce both crops. While the pattern is less extreme, it is still clear that cassava is most
important in the North and along the coastal areas, while maize is most important in the

central and interior areas.

‘

5 In Mozambique, the unit of display is the estrato, which is a conglomeration of several
districts with similar agro-ecological characteristics. This is the smallest unit available
for displaying TIA data accurately.

29

3. MARKET PARTICIPATION

3.1 Importance of and Factors Affecting Market Participation

The population of Zambia and Mozambique has been Shifting from farms to the cities for
some time (Figure 10). While the proportion of Zambia’s population residing in urban
areas has been well underway since before Independence, in Mozambique the trend
began in the late 1970’s. In both countries, an urbanizing population puts pressure on
national food systems (Tschirley, 2006). Governments are challenged to help spur the
marketing of local food production to keep pace with urban demand. Increasing the
number of households who can profitably commercialize their agricultural goods by
identifying circumstances in which farmers produce and sell marketable surplus is central

to this goal.

Many rural households, particularly the poorest, have few opportunities for off-farm
employment. Selling agricultural surplus is often the best avenue for acquiring necessary
goods and services. The poorest tercile of households in Zambia gain only about 5% of
gross household income from off-farm sources and the second tercile gains less than
25%. This group, which represents nearly three-quarters of the smallholder farming

population, gains the remaining income from agriculture (Zulu, 2007).

30

Figure 10: Zambia and Mozambique, percent of pepulation living in urban areas

 

Urban Population (% of Total)

45.0
40.0
35.0
30.0
25.0
20.0

15.0
10.0 —Zambia
5‘0 — — Mozambique

‘
’——

 

1 950 1 96 0 1970 1980 1990 2000 2005

 

 

 

Source: http://globalis.gvu.unu.edu/

For households to realize significant gains from trade in agriculture, they must transition
from subsistence production to specialization, honing their comparative advantage by
producing fewer crops in order to realize welfare gains from trade. As household welfare
increases, demand for off farm goods and services increases; as farming households
increase demand, it stimulates non-farm sectors of the economy. Whereas subsistence
households must provide the majority of inputs and products themselves, with properly
firnctioning markets and increased specialization, products can increasingly be purchased

and sold in markets (Boughton, 2007).

While the goal of increased commercialization has been in sight for some time, attempts
to reach it have encountered several pitfalls. The first of these became known as the
“food price dilemma”. Following independence, many Afiican leaders believed that

policies designed to “get prices right” were the key to increasing food supply and would

31

eventually lead to national food self-sufficiency and economic growth. In the 1980s it
was thought that if food prices were sufficiently high, rural producers would respond by
increasing output which would meet the needs of the growing urban population (Jayne,
1999). This attempt to stimulate production through high prices simultaneously put
pressure on consumers in the retail market. Moreover, in spite of the higher prices,
farmers did not expand supply as expected. It quickly became clear that prices are

necessary but not sufficient to stimulate and sustain the desired supply response.

By the end of the 1980’s, little improvement had been registered from pricing policies
alone. It was recognized that household-specific market failures in food and labor led a
large proportion of smallholder farmers to rationally choose self-sufficiency over
commercialization. In a general sense, most markets exist, but the costs of entering the
market may be higher than the gains from doing so, leading some households to
participate in the market, while some do not. (de J anvry, 1991). Transport costs,
information costs, and the opportunity cost of time are examples of obstacles to market
access that can prohibit market participation. Additionally, markets can be
geographically isolated and only partially competitive. Farmers face prices that are
volatile fiom year to year and highly negatively correlated with their own production.
Additionally, the large wedge between producer and retail prices leaves only the
wealthiest farmers in a position to enter the market on a sustained basis (Fafcharnps,
1992). The higher the transactions costs, the riskier the transaction and the more likely

households are to refrain from trading.

32

Recent studies have focused on market participation as a function of available household
assets (Boughton, 2007) building on the idea that a household must meet a minimum
threshold to be able to participate in the market, otherwise the household falls into a “low
equilibrium poverty trap” (Barrett, 2002). Small-scale farmers in rural Afiica face a
number of non-price related constraints to market participation. Specific determinants
such as land access, animal ownership, farm size, access to information, and availability

of labor all impact market participation.

Households need access to improved production technologies and to public and private
assets to successfully market foodstuffs. Omitting any one of these key elements is
sufficient to prevent a household from responding to otherwise favorable market
conditions (Barrett, 2007). Productive technologies such as improved seed and
husbandry techniques and improvements in public goods such as infiastructure and

market information sharing are important factors in market participation.

3.2 Summary of Maize Sales in Zambia and Mozambique

The previous section addressed the importance of marketing for smallholder welfare and
economic growth. It was noted that there are several factors that must be accounted for
before a household will enter into a crop market. This section smnmarizes the maize
selling trends in Zambia and Mozambique, both nationally, and at the AEZ level to

demonstrate the unequal distribution of maize selling both within wealth groups and

geographically.

33

3.2.1 Zambia

In Zambia, the agricultural sector is the main livelihood for more than 60% of the
population. Agriculture accounts for approximately 20% of GDP, and employs nearly
two-thirds of the labor force (FAO, 2006). Although maize is a very important cash crop
for Zambian households, the distribution of maize sales is highly skewed. In Table 5,
farming households are divided into mutually exclusive groups based on maize
production and marketing behavior: households that do not grow maize, households who
grow maize but do not sell and households that sell maize separated by quartiles of

quantity sold.

Nearly three quarters of smallholder farmers in Zambia produce maize, but only 35-3 7%
of maize producing households actually sells any maize. It is immediately clear that of
the households who sell, a very small percentage is responsible for the lion’s share of
quantity sold. Seven percent of households for each year of data for Zambia are
responsible for about 80% of the total marketed supply of maize. In a recent study
focused on the Zambian maize market, Zulu et a1 (2007) noted that the top 5% of maize
selling households enjoyed incomes of 8 to 9.2 times those of the non-maize selling
group. The welfare gains from maize market participation, then, accrue to a very small

propOrtion of the smallholder farming sector.

34

Table 5: Maize marketing in Zambia and Mozambique

 

 

 

(1) ZAM 1999/00 ZAM 2002/03 MOZ 2001/02 MOZ 2004/05
(2) (a) (b) 18) AIR (8) (b) (a) (b)
(3) Grew no maize 25% - 19% - 20% - 22% -
(4) Made no sales 47% - 52% - 59% - 60% -
Maize Selling Households
(5) Quartile 1 7% 3.5 7% 6.5 5% 4.1 4% 3.3
(6) Quartile 2 8% 20.0 7% 21.0 5% 10.5 4% 7.8
(7) Quartile 3 6% 39.2 7% 47.6 7% 34.0 6% 31.5
(8) Quartile 4 7% 210.0 7% 295.0 5% 204.0 4% 97.4
(9) Total 100% 272.7 100% 370.1 100% 252.6 100% 140.0

 

(1) Country and survey year to which the data refers

(2) Column (a) is the percentage of the population who participated 'n the activities below.
Column (b) Is the total maize sales in ('000 MT).

(3) Percentage of small-scale farming households in each country who did not produce maize.

(4) Percentage of households who did produce some maize. but sold none of their production.

(5) - (8) Quartiles of maize sales. Quartile 1 is smallest, Quartile 4 is highest

The spatial distribution of maize sales in Zambia is also highly concentrated.
Disaggregated by agro-ecological zone, the percent of households selling maize and the

. percent of total sales is clearly highest in AEZ 2a. This zone is fairly well connected in
terms of infrastructure. The bulk of the maize sales take place near the major north-south
roadway and the eastem-most highway. In these areas, maize production has been highly

encouraged since colonial times and farmers have a culture of producing maize for sale.

While AEZ 2a is clearly the most active in terms of total sales volume, it is not the zone
with the highest proportion of maize producers who sell. In both years, a higher
percentage of households in AEZ 3a (40 — 45%) grew and sold maize than in any other
zone. It appears that households near the mining areas and cities of AEZ 3a make maize
marketing decisions differently than households in AEZ 3b. This helps support the
observation that infrastructure and proximity to markets plays an important role in maize

marketing in Zambia.

35

Table 6: Zambia spatial distribution of maize sales

 

 

 

 

 

ZAMBIA
1 999/00 1 2 3 4 5
AEZ 1 9% 63% 28% 4% 10.1
AEZ 23 3% 62% 35% 67% 184.0
AEZ 2b 20% 64% 17% 2% 6.6
AEZ 3a 18% 42% 40% 17% 47.3
AEZ 3b 57% 24% 18% 9% 25.3
Total 24% 48% 28% 273.3
2002/03
AEZ 1 7% 78% 15% 3% 9.8
AEZ 2a 3% 69% 28% 62 % 228.0
AEZ 2b 1 6% 65% 19% 2% 5.8
AEZ 3a 1 4% 41% 45% 20 % 75.4
AEZ 3b 47% 26% 26% 14% 51.3
Total 1 9% 52% 28% 370 .3
Column: 1: Percent ofall HH that did not produce maize

2: Percent of all HH that produced maize, but made zero sales

3: Percent of all HH that produced and sold maize

4: Percent of total sales

5: Total Sales('000 MT)

On the Opposite end of the Spectrum, households in AEZ 1 and AEZ 2b are among the
most likely to produce maize (> 63% and > 64%, respectively), but are responsible for
the smallest quantities of marketed. Distance from urban markets is partly responsible
for this trend, but it is also true that households in these areas, which are less productive
than AEZ 2a, suffer from relatively meager harvests, leaving little surplus for sale. AEZ
1 is characterized by the lowest rainfall, and AEZ 2b is known for sandy soils that are not

well suited to maize production.

Figure 11 classifies and maps the Primary Sampling Units (PSUs) in the two countries

(Survey Enumeration Areas (SEAS) for Zambia, Unidade Primaria de Amostragem UPA

for Mozambique) into quartiles of total maize sales. The circle sizes represent the total

36

sales quartile classification of each unit; these are country specific and do not represent
absolute magnitudes of sales between countries. The map is intended to highlight the

geographic dispersion of maize sales within each country.

In Zambia, the shaded circles represent the top 25% of maize selling SEAS between the
1999/00 and 2003/04 seasons. Likewise, shaded circles in Mozambique represent the top
25% of maize selling households at the UPA level for the 2001/02 and 2004/05 seasons.

Because neither survey has accurate spatial locations for households, the next-smallest

. . 6
umtrsused.

The distribution of the tOp 25% of maize sales in Zambia is very concentrated along the
north-south highway which stretches from the border with Zimbabwe in the South
through Lusaka to the Copperbelt on the Congo border. Other obvious clusters are in
productive and populous Eastern Province bordering Malawi and Mozambique and near
Lake Tanganyika in Northern Province. The patterns in Mozambique are discussed in the

following subsection.

 

6 PSU centroids are used to locate the data based on latitude and longitude information
collected during the survey. In Mozambique, many UPAS were recorded without spatial
information. For display, UPAs without spatial reference are located at the district-level
centroid. For this reason, a number of circles in Figure 11 have a “bull’s eye” pattern
where data for more than one UPA overlap.

37

 

Figure 11: Distribution of maize sales by quartile
'34.. - ~ » 5,4

LI!
A.
3;:
2:: ‘1
'7 Le

ti \\,.,J- . “.

Zimbabwe

South Africa

       
 

  

 

Maize Sales
by Quartile
Zero Sales

 

 
 

> p25 < p50

0

> p50 < p75
0 Sell > p75

0

 

 

  

w ‘~_-~‘.-‘

 

/

 

 

Source: Zambia Supplemental Surveys & Mozambique TIA 2002, 2005

38

3.2.2 Mozambique

As in Zambia, it is immediately apparent from Table 5 that the distribution of maize sales
in Mozambique is highly concentrated. In both years for Mozambique, about 60% of
farming households produced maize but did not sell any. Just over 20% of farming
households sell all the marketed maize in the country. Of these households, the top five

percent is responsible for about three quarters of the marketed volumes.

The geographic distribution of maize sales is also highly concentrated, with the majority
of marketed volumes taking place in AEZ 3 (Table 7). In both years, nearly two thirds of
marketed volumes originated from this region. This is likely a consequence of
dependable rainfall, a large population of farming households, and reliable demand from
the urban areas of Southern Malawi. What is particularly interesting about AEZ 3 ’3
performance compared to the other zones is that in contrast to the other two zones,
households in AEZ 3 marketed a higher percentage of maize in the 2004/05 than in
2001/02. The 2004/05 growing season was a below-average in terms of rainfall, and
many farming households suffered significant crop loss, even in AEZ 3. Because AEZ 3
is known to be a cassava and maize producing region, this qualitatively supports the
proposition that households producing both crops are able to market more maize than
households without cassava. However, because a high percentage of households in all

three zones produce maize and cassava, further analysis is necessary.

39

Table 7: Mozambique spatial distribution of sales

 

 

 

 

 

M OZAM BIQU E

2001/02 1 2 3 4 5
AEZ 1 ‘ 15% 77% 8% 10% 24.8
AEZ 2 15% 68% 1 6% 34% 87.3
AEZ 3 24% 49% 27% 56% 141.0
Total 20% 59 °/o 21% 1 00% 253.1
2004/05
AEZ1 15% 80 % 5% 3% 3.5
AEZ 2 19% 68 °/o 1 3% 28% 39 .9
AEZ 3 25% 53% 22% 69% 96.6
Total 22% 60 % 1 8% 1 00% 1 40 .0
Column: : Percent of all HH that did not produce maize

1
2: Percent of all HH that produced maize, but made zero sales
3: Percent of all HH that produced and sold maize

4: Percent of total sales

5: Total Sales ('000 MT)

In addition to selling the largest volumes of maize, households in AEZ 3 are the most
likely to grow maize for sale. AEZ l is the least likely to sell, with less than 10% of
households entering the market in either survey year, followed by AEZ 2 in which

approximately 15% of households sold.
3.3 Factors that impede maize sales

3.3.1 Physical factors

Farming households in Zambia and Mozambique face similar constraints to maize
marketing. Chief among these constraints is the inability to satisfy food consumption
needs. The previous section called attention to the fact that a small number of maize
sellers account for the majority of marketed volumes. These households have identified
ways to satisfy food requirements either through other income streams which allow them
to purchase adequate food, or because they have the ability to produce adequate stocks to

satisfy consumption and to be available for sale.

40

Poor infrastructure and long distances to market raise the cost of selling maize and deter
households from marketing their surplus. Even if a household has a surplus, the benefit
from selling may be less than the cost of making the sale. Included in these transactions
costs attributable to distance are price discovery costs and the opportunity cost of time.

Figure 12 illustrates the vastness of the region and the relatively sparse road network that

links the major cities with the rural production areas7.

Finally, uncompetitive markets for maize can prevent households fi'om profitably selling
their surplus. If markets are not competitive, traders are able to purchase maize at a low
price. This is fundamentally the same as adding a cost to the maize selling household
which is equal to or greater than the benefit from seeking out a trader willing to purchase

at a higher price.

 

7 Portions of the road network in Figure 12 are more poorly maintained than they appear.

41

 

Figure 12

Road network and population dens

Sconce...

 

 

 

 

881-
80.8.8.
885'
8.:'
2-8
mom

«is. can can

  

 

 

 

   

 

 

CIESN/CIAT, 2005

Source

42

3.3.2 Political factors

This section addresses the policy environment which, like the physical environment, has

important consequences for improved market participation.

Zambia

The Government of Zambia continues to be a major player in maize production and
marketing. Despite reducing govermnent involvement in these sectors since the onset of
structural adjustment policies in the 1990’s, much of Zambia’s agricultural budget
continues to be funneled into the Fertilizer Support Program (F SP) and the Food Reserve
Agency (FRA). In 2005, these two line items accounted for over half of the
govemment’s expenditures on agriculture. In contrast, the proportion allotted to
agriculture infrastructure and irrigation development was 2% and 3%, respectively
(Jayne, 2007). This skewed resource allocation indicates that the government continues
to rely on Old habits rather than exploring new solutions. Unfortunately, both the FSP
and the FRA have the potential to reduce private sector involvement in input delivery and

maize marketing.

Rather than. supporting private trade in Zambia, the government influences market prices
and supply through the operations of the FRA. Often, government buying agents offer to
purchase maize at above-market prices. Doing so limits private sector involvement and
can eliminate incentive to export. Moreover, FRA agents can be inefficient and are

frequently low on cash which can lead to delays between transfers of maize to receipt of

43

payment (Whiteside, 2003). The FRA is a drain on public resources and if inefficiently

managed can harm small-scale farmers and private traders alike.

Another aspect of government policy that has potentially negative effects on smallholder
farmer welfare stems from a fundamental misunderstanding of the market behavior of
rural farm households. Trade restrictions on maize imports are frequently imposed. The
restrictions are founded on the belief that the majority of smallholder farmers derive
income from maize sales. Therefore the higher prices caused by restricting imports is
expected to lead to expanded domestic maize marketing which in tern improves
smallholder welfare. These restrictions continue despite recent studies that have revealed
that the majority of smallholder farmers, particularly the poorest group, are actually not
buyers of maize. Net buyers purchase more maize during the marketing season than they
sell, if they sell at all. Thus, higher prices may reduce the welfare of most smallholders

(Zulu, 2007).

Conversely, Zambia imposes export restrictions on maize grain in an effort to retain
maize for deficit areas of the country during years of below-average national production.
Eastern province is a highly productive maize area for which Lilongwe, Malawi is a
natural market. Export restrictions in the form of fees, licenses, or outright bans limit the
ability of farmers in Eastern Province to meet the demand. However, export restrictions
are only effective at stemming formal trade flows while the informal channels remain
active in the absence of enforcement. Even in the absence of outright export bans, small-

scale traders may eschew formal trade channels in an effort to avoid the delays stemming

from acquisition of required permits for formal trade that can only be obtained in the
provincial capital. It is thought that up to 80% of maize exported to Malawi is done

informally as a consequence (Whiteside, 2003).

Mozgmbique

 

The private sector in Mozambique has been able to respond to the relatively open policy
environment and has succeeded in linking surplus and deficit areas. Cross border trade is

important both for small holder farmers and urban consumers in Mozambique.

The urban areas of Southern Mozambique are set within rural areas known for marginal
maize production. Unable to address the growing urban maize demand from local
production alone, the southern cities rely partly on the central region which in most years
is capable of supplying surplus production. However, the central region alone can not
satisfy the need for maize, and therefore relies heavily on imports of maize grain from
South Afiica. Although maize is routinely imported, regulations greatly benefit large
millers in Maputo/Matola. Large millers are able to import maize grain, paying only a
minimal fee provided that the maize is milled once in Mozambique. Import dues for
imported maize meal and value-added tax for imported maize grain that is not

subsequently milled are significantly higher (Whiteside, 2003).

Maize is a tremendously important cash crop for small-scale farmers in northern
Mozambique. The highly populated southern region of Malawi is surrounded by maize
producing areas of Mozambique and is the natural market for Mozambican maize

surplus. Both formal and informal trade is active in the area. While it remains difficult

45

to calculate accurately, Whiteside (2003) estimated that in 2002, which was a year of
high demand in Malawi and normal production in Mozambique, over 300,000 MT of
maize crossed into Malawi through formal channels, and almost 225,000 crossed through

informal channels.

Figure 13 illustrates some of the regional maize trade patterns in the Southern Afiican
region and illustrates the importance of trade for Mozambique in particular. Often, areas
of surplus production are separated from their logical market by international boundaries.
To the extent that maize can flow across these boundaries unimpeded, both producers and
consumers can benefit. Highly productive areas such as Northern Mozambique benefit
from unimpeded access to the nearby urban consumption centers in Malawi (Haggblade,

2008)

46

rica

Trade flows, Southern Af

Figure 13

 

    

86 Loss.
250E .655. 8

.<.w.m

         

m>>m<ms=N

 
   

 

O <_m§<N
9355. 9.83..
035.
m30.m2<NO_2
£935.33
V , Odd

<_Z<NZ<._.

 

 

2008

Adapted from Haggblade,

Source

47

3.4 Potential consumption shock absorbers

In the urban setting, low-income consumers spend a large proportion of their incomes on
maize and maize meal making them vulnerable to the price spikes that accompany poor
national maize harvests. Because incomes are inadequate, households often reduce
consumption to unacceptable levels during price spikes, a situation that impacts the

health of household members.

In rural areas, farming households rely on a combination of own production and
purchases to meet their needs. Because the majority of small-scale farmers are net buyers
of maize in both Zambia and Mozambique, the availability of affordable maize grain and
maize meal is also important to them. For many households, especially in Zambia, maize
is the only staple food produced in adequate quantities to meet the bulk of the
household’s calorie needs. If production is inadequate, households rely on markets if
possible to fill the consumption gap, but if prices are out of reach — and if they have no

backup crop to rely on — they have little choice but to reduce consumption.

Long-term national food security requires access and availability of maize for both rural
and urban households. Currently, Zambia and Mozambique are plagued with periodic
production and price fluctuations which in turn lead to consumption volatility. At its
essence, reducing consumption volatility implies maintaining stable retail prices that are
within the reach of consumers, and ensuring that rural households are able to meet their

own consumption needs either through own production or purchase. Reducing

48

consumption volatility improves food security; there are at least two possible ways to

accomplish this goal: open trade and the widespread production of alternate food crops.

As we have seen in the previous section, Zambia and Mozambique have a mixed record
when it comes to open trade. Zambia’s ad-hoc approach to trade policy — barring imports
in an attempt to boost production and barring exports to retain food stocks — is less
flexible than Mozambique’s. AS a consequence, Zambia is unable to react to periodic
production deficits and therefore faces chronic food insecurity in certain rural areas and
highly variable maize meal prices in urban areas. Mozambique relies on imports to feed
the southern cities, but does not always allow free movement of commodities across the

border to Malawi, compromising the rural incomes of Northern Mozarnbicans.

Promoting the production of alternative food staples to maize could potentially act as a
consumption smoothing mechanism as well. Planting crops that are substitutes for one
another in consumption, but that do not have highly correlated yields, is ideal. Cassava
and maize are an example of crop diversification that has the potential to smooth
household consumption. Because cassava is a root crop that is resistant to rainfall
fluctuations, it can be left in the ground up to four years and harvested when maize crops

are severely depleted (N weke, 2002).

The ability of cassava to produce a reliable harvest during below-average maize
production years has a two-fold benefit to households who produce both crops. First, it

enables households to maintain a reasonable level of staple food consumption in their

49

own households. If maize is sufficient to meet the need, they can consume maize, but if
it is not, they can substitute maize with cassava. In Zambia and Mozambique, the
households that grow both staples likely consume some portion of both crops in a normal
year, depending on their preferences, and increase the ratio of cassava to maize in years

when the maize harvest is poor.

The second potential benefit of growing cassava in addition to maize is the flexibility to
respond to high maize prices. One persistent problem for most smallholder farmers is the
high inverse correlation between their maize harvests and the farmgate prices for maize.
In bad maize harvest years, when maize growing households ordinarily have no surplus
to sell, maize prices are attractive. Conversely, when maize production is high, and
households have surplus to sell, prices are low. If households produce only maize, they
have little flexibility to respond when prices are high and little incentive to respond when
prices are low. If, however, the household produces cassava in addition to maize, they
can shift their consumption bundles toward cassava and sell maize during times of high
maize prices, thereby supplying the market and gaining cash revenue without reducing

consumption.

50

4. DATA & SAMPLING METHODS

4.1 Zambia Supplemental Survey to the Post Harvest Survey

Since 1991, Zambia has conducted a nationally representative Post Harvest Survey
administered by Zambia Central Statistical Office (CS0) and the Ministry of Agriculture
and Cooperatives (MACO). Additionally, two supplemental surveys to the PHS were

conducted for harvests in 2000 and 2003. In both the PHS and the supplemental surveys,

all districts of Zambia were sampledg. Within each district, a random sample of survey
clusters were identified, from which households were randomly selected. Small and
medium scale farm households — the focus of this study — were interviewed separately
from large scale households. For convenience throughout the paper, small and medium

scale households are referred to together as the “smallholder” sector.

The sampling frame for the PHSS conducted between 1991 and 2000 is based on the
Census Supervisory Areas (CSA) and Survey Enumeration Areas (SEA) defined using
the 1990 census. Changes in population distribution have led to the necessity of
recalibration the sample frame following the completion of the 2000 census, and was
implemented beginning in 2004. The data collected prior to 2004 is still nationally
representative sample of households, but it is less efficient than the latest sampling frame

(Megill, 2004).

 

8 54 districts fiorn 1991 — 1996, and 72 districts thereafter.

51

The supplemental surveys to the PHS were conducted in coordination with C80, MACO,
and Michigan State University (MSU). In the first supplemental survey, conducted in
May 2001, households sampled in the 1999/2000 PHS were re-interviewed and asked
additional questions not included in the original PHS in order to enrich the scope of the
dataset. The original PHS omitted several important variables (including cassava
production information) that were captured by the supplemental survey. The first
supplemental survey successfully re-interviewed 6922 households out of 7694 from the
99/2000 PHS, and asked them to recall production and sales information from the
1990/2000 cropping year. The second supplemental survey successfully re-interviewed
5344 of the first supplemental survey’s households, asking them to recall information
from the 2002/2003 harvest season. Together with the first supplemental survey, this
established the first household-level panel dataset for smallholder farmer harvest

information in Zambia.

4.2 Mozambique National Agricultural Household Survey (TIA)

Data utilized in the analysis of Mozambique is drawn from the national household
agricultural survey known as the T rabalho do Inquerito Agricola (TIA). Collected
through the Mozambique Ministry of Agriculture, the TIA provides nationally
representative panel data on household level production and marketing of food and non-
food crops. The 2002 TIA sampled approximately 4900 randomly selected households
living in 80 out of 120 rural districts in Mozambique and subsequent years are more
comprehensive. The TIA is a recall survey carried out after the peak crop marketing

season and focuses on collecting detailed information on crop production and marketing

52

and household incomes. This study utilizes data fi'om two representative years of the

TIA: TIA 2002 and TIA 2005.

The limitations of the TIA are similar to those of the Zambia supplemental surveys; the
TIA relies on farmers to recall production and marketing information from earlier in the
year, reducing to some degree the accuracy of the information. Additionally, because the
TIA does not collect information from every district in Mozambique, the ability to
extrapolate to the national level is perhaps more difficult. However, the data available in
the TIA has several advantages over the Zambia PHS. The survey instrument of the TLA
has been largely unchanged since TIA 02, and has includes a full complement of relevant
variables. Additionally, the accuracy of the crop information is enhanced by a system of
data collection that measures a sample of fields to adjust household recall of hectares
planted, and helps farmers accurately determine the share of a field planted to a crop by
separating colored beans according to intercropped ratios. The TIA, like the
supplemental surveys, is the most accurate panel information available for analysis of this

kind.

4.3 Geospatial Data

This study makes extensive use of Geographic Information Systems and spatial data to
put the household level production and marketing information in a geographic context.
The types and sources of geographic data are diverse and have only been easily
accessible to researchers outside of geography departments for the past few years. The
ability to interact spatial information with information available in the nationally

representative agricultural surveys is a key feature of the empirical analysis in this study.

53

Geographic datasets can be overlapped with other types of data to create new
information. The agricultural production surveys carried out in Zambia and Mozambique
all contain geographic information that is largely underutilized in ordinary analysis. In
both Zambia and Mozambique, information on the location of households from the
production surveys can be approximated to sub-district level. Matching the information
contained in the surveys with the geographic data created by other means enables us to
add geographic data relevant to household records that were not originally collected in

the surveys.

Two key pieces of geographic information are utilized in this study: a) Euclidian distance
from Primary Survey Unit to primary road and b) seasonally accumulated and long-term
average rainfall. For distance to primary roads, data is created using GIS software to
compute the straight-line distance to primary roads from the approximate locations of
households. This distance is converted to kilometers and the variable is then appended to

the dataset for each year and incorporated into the statistical analysis.

For seasonal rainfall, monthly precipitation grids from NASA’S Tropical Rainfall

Measuring Mission (TRMM)9 are summed from November to May, the approximate
maize growing season in Southern Afiica. By overlaying the resulting seasonal rainfall

surface atop the spatially located Primary Survey Units in Zambia and Mozambique, a

 

9 TRMM data used in this study were acquired using the GES-DISC Interactive Online
Visualization and Analysis Infiastructure (Giovanni) as part of the NASA'S Goddard
Earth Sciences (GES) Data and Information Services Center (DISC).

54

continuous variable for rainfall is created. Using the same TRMM data, the standard

deviation of rainfall is calculated.

Similarly, using a combination of geospatial datasets, comparable agro-ecological zones
are delineated as described in chapter 2 for both Zambia and Mozambique based on a
high-resolution, interpolated, 30-year average rainfall model fi'om WorldClim (see

Hijmans, 2005 for description of WorldClim database).

55

5: ANALYTICAL APPROACH AND HYPOTHESIS

The unit of interest to this study is the small-scale farming household that relies primarily
on own maize production to meet household staple food needs and to obtain non-farm
goods and services via cash from sales. This section describes the household
environment and sequential nature of household decision making that leads to the

hypothesis and econometric models of subsequent sections.

5.]: Household decision making

At the onset of the rainy season, the household determines how much maize and other
crops to sow given available labor, land, and productive assets, but is unable to know
exactly what the maize yield will be at harvest time. Throughout the six month maize
growing season, the household attempts to maximize its chances of harvesting an
adequate crop given factors Such as illness, labor availability, and farming skill, but
unpredictable rainfall generates widely variable harvests from year to year. Only once
the crop is harvested can the household make consumption decisions and begin to

consider marketing available surplus.

Maize marketing is a two-step decision in which households first decide whether to enter
the market, and then decide how much'of their crop to sell. A number of factors
influence these decisions such as available surplus, price, transportation costs,
information costs and negotiation ability. If a household produces only maize, the
available surplus equals the total harvest minus household consumption. For many, this

will be a very small quantity and due to varying market access, lack of necessary assets

56

or high transactions costs, some households may rationally choose to abstain from market
participation altogether, remaining in autarky (Boughton, 2007). If, however, the
household produces cassava as a substitute for maize in consumption, the quantity of
maize available for sale can potentially equal the quantity of cassava consumed. In an
extreme scenario a household with both crOps could consume cassava as the sole staple,

reserving the entire maize harvest of for sale.

The nature of cassava, which continues to increase in volume in the ground and can be
harvested anytime between two and four years after planting, makes it a reliable
storehouse of calories for the household, and can act as insurance for times in which the
maize harvest fails. Collinson (1983, p. 26) describes the household’s strategic use of

cassava as a substitute for other crops using an example from Tanzania:

“Particular crops are often grown especially for insurance. Cassava is a
case in point with the Sukuma [Tribe]. It is not a preferred starch staple,
but its high productivity in terms of bulk and its capacity for storage in situ
in the ground for up to three years make it an ideal famine reserve crop.
The aim of the farmer will be to have cassava available and mature in
future periods of expected food crisis. Cassava requires little work after
being weeded at the beginning of its second season. In Sukumaland the
labor required for planting is minimized by intercropping and by planting
the cassava at a time when there are few other demands on labor. This late
planting means the sacrifice of optimal yields, but it still gives five to ten
times the bulk of grain crops per unit of labor used. When cassava
supplies are low after a famine season, reestablishrnent may take first
priority when the rains arrive. The amount of mature cassava likely to be
available will influence decisions on the urgency of staple grain supply
and contingency plantings of late grains in a poor season. It will thus be
important in the allocation of resources to other crops.”

Because an investment in cassava gives households the ability to smooth staple

consumption, it can also provide them more flexibility with regard to maize sales than

57

households who produce maize only. This is the foundation on which the hypothesis

below is built.

5. 2: Hypothesis

Given the relationship between maize and cassava, and based on the previous chapters’
discussion of their importance to smallholder farmers, the hypothesis of interest concerns
the relationship between cassava production and maize sales. Explicitly, the hypothesis

to test is that maize sales are positively related to cassava area planted.

5.3: Key variables for hypothesis testing
Equation 5.1 describes the household’s maize marketing decision. Kilograms of maize

sold, Smi, defined by the quantity of maize harvested, Qmi, minus household
consumption, Cmi, is hypothesized to be a function of household maize harvest, Qmi,
maize price Pmi, cassava area Kci, geographic variables gi, demographic variables di,

wealth and income variables, yi, and productive assets, ai.

(5.1) 3111i: Qmi - Cmi =f(Qmi, Pmi, KCi, 88 di, yi. 3i)

Standard factors influencing maize sales include kilograms of maize harvested, (Qmi)
and price of maize (Pmi). Including these variables in the econometric modeling is

important to evaluate the impact of cassava hectares (Kci) in a ceteris paribus context.

Questions of endogeneity stemming fi'om the inclusion of maize production as an

58

explanatory variable are resolved by the above discussion of the sequential nature of the
production and sales decisions. Because production and sales decisions do not take place

simultaneously, there is no endogeneity bias.

The geographic variables that make up (g) in Equation 5.1 include Euclidian distance to

primary road, and seasonal mean and standard deviation of rainfall. These variables are
included to hold constant aspects related to marketing that have a strong regional

connection and are likely to influence the cost to marketing goods. Demographic

variables (di) include the age and gender of the household head and the number of

members in the household. The latter variable is an indication of available household
labor and consumption levels. The former two variables are expected to influence

household transactions costs and flexibility to respond to market signals. Income
variables (yi) include salaries, ownership of businesses and remittances”). These
variables are thought to change a household’s incentives to participate in maize markets
by giving them other income earning alternatives. Finally, productive assets (ai) include

total land available to the household, cattle ownership, and transport and radio ownership.
These variables influence a household’s ability to substitute income, increase maize

production, or reduce search and other transactions costs.

 

'0 These variables are only available for the analysis in Mozambique.

59

6. EMPIRICAL ANALYSIS

6.1 Estimation Methods for Limited Dependent Variable Models

Limited Dependent Variable models are utilized when the dependent variable is
substantially restricted and does not follow an approximately normal distribution.
Because LDVS violate the normality assumption, they lead to biased and inconsistent
estimations in OLS. Smallholder maize market participation is an example of a specific
LDV called a comer solution problem. Distributions of this type are continuous over
some range of observations, but “pile up” at zero (Wooldridge, 2002). Optimizing
behavior for maize sales leads to a roughly continuous distribution over the subset of
households who sell, but for the larger portion of households who do not sell, zero

kilograms sold are observed.

Corner solutions have traditionally been dealt with using the Tobit model, but there are
other, less restrictive limited dependent variable models suitable for this type of analysis.
This section briefly discusses the Tobit model in the context of household maize
marketing and introduces two alternative limited dependent variable models: the

Heckman selection model and Cragg’s Tobit alternative model.

The Tobit model (Tobin 1956) is familiar to economists seeking to model behavior in
which the dependent variable takes on an approximately continuous distribution after
being censored at some (usually zero) point. There are a number of examples of so-

called “comer solution” problems in which a non-trivial proportion of the population

chooses a zero quantity or dollar value of something. Alcohol consumption and annual

6O

family charitable contributions are two familiar examples in which there is likely to be a
large proportion of the population who do not participate in the activity, and the
remaining portion of the population participates over a wide range of values

(W ooldridge, 2006).
The Tobit model can be expressed by the following latent variable model:

(6.1) y* = x [3 + u, u|x ~ Normal (0,62)

(62) y = max (081*)

where y equals the latent variable y* when y* 2 0 and y = 0 when y* < 0. Because y* is

a latent variable that follows a roughly continuous distribution, y must have a continuous

distribution over strictly positive values.

For the purposes of this paper, the unconditional expectation E(y|x) is of particular

interest. For the Tobit model, this can be expressed by the following:
(6.3) E(y|X) = (FOX 0 /6)X II + 090‘ II /0)

The first term refers to the probability that y > OIX, and the second refers to E(y|y > O,
x). The second term incorporates the inverse Mills ratio which indicates that the

expected value of y > O is conditional on Xfl plus a positive term. The equation indicates

61

that when y follows a Tobit model, E(y|x) is a nonlinear function of X and B. The '
following expression illustrates the method for computing partial effects of a change in

Xj:

(6.4) dE(y|x) / dxj = flj (I)(x 8/ 0)

where the coefficient, (I), is the cumulative probability of x Bl O'.

The Tobit model estimates the probability of an observation being positive at the same
time as it estimates the magnitude of the response. The model assumes that the effect of
the explanatory variables responsible for the actor’s participation decision (y= 0 versus y
> 0) must also be responsible for the quantity decision (the magnitude of y when it is
positive), and that they affect both decisions the same way, resulting in a single set of
coefficients. In some cases, however, requiring each stage to be estimated

simultaneously may not accurately reflect the decision making process.

In the discussion of household market participation, the debate over the Simultaneous
nature of the decision making process has been challenged by some researchers.

Bellamere and Barrett (2006) described the situation this way:

“If households make participation and volume decisions
simultaneously, they effectively pre-commit to a volume before
learning information available to them only once they arrive at market.
This ex ante decision making effectively gives the traders with whom
the household interacts market power by rendering the household’s
demand (supply) inelastic with respect to new market information (e. g.,

62

prices) they discover, leaving poor pre-committed households

vulnerable to exploitation by astute traders. If, however, households

make marketing decisions sequentially, then they retain greater

flexibility once they arrive in a market, making their purchases or sales

volume decisions ex post based on new information discovered at

market, thereby reducing traders’ capacity to extract much or all the

gains fi'om trade. (pp 324)”
One model that explicitly separates the participation decision and the quantity or amount
decision is the Heckman selection model. Heckman (1979) proposed an estimation
technique that addresses the problem of zero observations generated by non-participation
due to respondent’s self-selection. The model uses a two-step procedure in which a full
sample Probit estimation is followed by a censored OLS estimation applied to the non-
zero subsarnple. The Heckman estimation technique is a generalized version of the Tobit

model that accounts for the possibility of omitted variables that influence both decisions

(W ooldridge, 2006; Wodajo, 2008).

The classic example of a self-selection problem for which the Heckman technique is used
concerns the wage offer for married women. In this example, there are two equations
that must be evaluated: (i) the probability that a married woman will enter the workforCe,
and, (ii) the observed wage conditional upon entering the workforce. The selection
problem arises because the wage offer is observable only for working women. In this
Situation, the analyst must correct for sample selection bias by evaluating the first part of
the question separately from the second. The Heckman model allows for the possibility
that the two decisions do not depend on the same set of factors and separately identifies
the affect of each factor on each decision. In the wage offer example, the likelihood of

entering the workforce may be based on income from the husband and the presence of

63

young children in the home, whereas the wage offer equation may be related to the
woman’s education level and experience. It iS clear that some variables have an impact
on the participation equation but are not appropriate for the wage offer equation

(Heckman, 1979, Wooldridge, 2006).

In Heckman’s two-step estimation method, the first step refers to the participation
(selection) equation and the second to the amount or quantity equation. The equation of

interest is the quantity decision:

(6.5) y = x8 + u, E(u|x) = 0
and the selection decision is:

(6.6) s= l[zy+VZO]

where S = 1 if y is observed, and zero otherwise. The elements of x and Z are always
observed, by assumption; [I is the coefficient on X, and y is the coefficient on 2. It is
assumed that the selection equation depends on observed values, 2 and unobserved error
V, and that z is exogenous to the continuous equation. The major assumption that
justifies the Heckman selection model is that the error terms of the two models, u; and Vi,

are correlated, and therefore must be taken into account.

Estimating the Heckman model in two steps involves a Probit for all n observations and a

censored regression on the positive observations that includes the inverse Mills ratio for

64

each observation as an additional independent variable. To account for the correlation
caused by the sample selection bias, the inverse Mills ratio is included as an additional

regressor in the second stage of the estimation.

The Probit model for the selection decision takes the following form:

(6.7) P(s = 1|z) = (on)
and

(6.8) E(ylz.s = 1) = x9 + 91121)

The equation above shows that the expected value of y given 2, and the observation of y
(that the selection decision is positive) is equal to XB plus p which is determined from the
inverse Mills ratio in the second equation. If p = 0, then OLS on the selected sample

consistently estimates B. But if p ¢ 0, it would be akin to running OLS on a model with a

missing variable, leading to bias (Wooldridge, 2006).

The third estimation technique considered here is the Cragg truncated normal hurdle
model. It is also an extension of the Tobit model in which the requirement for
simultaneous selection and quantity decisions are relaxed (Cragg 1971). The technique
combines the likelihood function of Probit (for the selection decision) and truncated

normal regression (for the continuous decision), but, unlike Heckman’s model, assumes

65

no correlation between the errors of the first and second equation, conditional on a
properly specified model (W ooldridge, 2006).

Cragg Truncated Normal Hurdle Model:

(6.9) w', = xii + u
yi=W*i ifSi= I andw*i>0
or y; = 0 else

This indicates that the observed quantity yi is non-zero when participation is recorded and
a positive quantity of y is observed. Cragg’s model is estimated through a combination
of a Probit function for the participation decision and truncated regression estimated with

maximum likelihood for quantity decision. One way to express this is: .

Expression for the participation decision follows a Probit:

(6.10) P(s = llx) = (I>(xy)

Continuous decision given y > 0 and x follows a truncated regression:

(6.11) flylx,y > 0) = I9<x13/o>1“<p[(y — x137 or o. y > 0

which leads to the density for y|x:

(6.12) flYIX) = [1 -<I><xr)1'”"°] 1960196176)" cply- XB/OI/9}1[Y> ”1

66

Unlike the Tobit model the Cragg model allows for the two parts of the estimation to be
estimated independently, and while the first hurdle ofCragg’S model is identical to that
of the Heckman model, the second stage is estimated with maximum likelihood rather
than OLS. An advantage to the Cragg model is that it is relatively straightforward to

calculate average partial effects of (ylx).

In summary, the Tobit model is well established, and partial effects are calculable, but the
model is limited by the restrictiveness of the assumption that the participation and
quantity decisions are based on the same mechanisms. The Heckman model is less
restrictive, and has the added benefit of taking into account unobserved variables that
affect both parts of the model, but interpretation of results is not straightforward, and
computing partial effects for comparison with the Tobit model is beyond the scope of this
paper. Finally, while the Cragg double hurdle model can be estimated jointly and partial
effects can be calculated, the assumption of strict independence between the participation

and quantity decisions is still open to debate.

6.2 Methods

6.2.1 Variables

The aim of the econometric techniques utilized in this thesis is to estimate the impact of
cassava hectares on kilograms of maize sold. To reach this goal, two of the models
require the specific identification of the participation variable. The Heckman model and
the Cragg double hurdle model both estimate a Probit model in the first stage; for the

question at hand the binary dependent variable = 1 if the household sells maize and = 0

67

otherwise. For the Tobit model, this variable is not explicitly utilized, but is implicit in
the technique. The dependent variable of the Tobit, and of the second stages of the

Heckman and Cragg model is kilograms of maize sold.

The cassava hectares variable is the independent variable of particular of interest to this
study. In order to evaluate it ceteris paribus, a number of other independent variables are
included that fall into the following broad groups: demographic variables, wealth and
income variables, and asset variables. Tables 8 and 9 provide a description of each .
variable, indicate the type of variable it is (continuous or binary), and display their
population weighted means. Due to the differences in survey design between Zambia and
Mozambique, not all variables are identical for the two countries. However, the available

variables that are most likely to have an affect on maize sales are included.

68

Table 8: Zambia explanatory variables and means

 

 

 

 

 

88.53 86 889.0 8o no» 2:8 98824.8: an».
99.6.0 8o Ru 58 So no» 9853 S 9.3 388: no
8888 8... 328.0 so 8» 883: o 23 588... 3:5
Saw; 8... 8888.0 9o no» oiononfio 5:88: 98
$598 to 8:3 13 mo» 82356 388: 98.
«man? «we 8am... 3.? oz 9288; 9 oasis 22b SHE: use 3.9
R289 8.2. 83th 89 oz 82. 8:83:62 8..
BESo 98 ESE Re no» 35826822912 582
E8.“ So was; So oz n85: oooonaonoEE one
.4989 8.3 853.9 8.6 oz __oto._o§oon__o Sensooozoofim En
sauna 8+9 84.8.8 8.8m oz inocafiooo Sign! ES:
838 ES $9.8 8.8 oz “896...... 82 one 288% Sofia. 9915.
98.8? 3.5 85.8 8.8m oz SE one: 65% 532 £88
Romano 898$... BERN 8.3384 oz 8585.822 Sabotage. «L213
82:: 8.82 8:8? 880.. oz conga; 8853259.: 522.9.
883 «no 2828 So oz 8838a? 38;
on so... on in: in 58:85 son-3

n. SE 8.82. (HEN

 

Source: Calculated from the data

69

Table 9: Mozambique explanatory variables and means

 

 

 

 

 

-wgd 8o 8» £393! canons: xouz

85:6 8.0 3853 8c 8» Eogooiooz :25
3380.0 So mammoth. So no» olesoEaaio 2238: 9.15
R395 no.0 3593 and 3» olefin 2.6 2238... 83
36.8% :58 Satan 2.»: oz alofiéooofigoum E!
ENS nag nag Sago oz @ooiébono v.8 3.3
83.2w» 8.5 3+8: 882 oz 383$ .35.» .3} use «2an
$2.8m 8.8 “.888 8.58 on» 3355382.. ofiouozugfisouw Ego
«353 SN Shara k.» oz unsung 22%.: ofizooofioofae. 3:9
589 No.9. 589.9 mm? oz 2.2 unions»? 0?
583.0 mg F310 03 on» one soon; 2288: 582
888+ Ro «85¢.» 8m oz obosusoafiaofobopsz one
.383 8.8 35.2 88 oz 15.2 EEEoBSfisuo 32.5 on
8.6.9.“ atoms 28.8w 58.. oz .25! 1888 oaazsafl 5:5
.533 :8 3.5.8 88 oz zeaoooyvoaooouessfizo Exam. 82:5
52% man gamma 9N oz 8: finesse“. 535. ESE
8mm? 8,855 Sig, 888%.» oz A8558?! onaeooosaoofi «1.219.
9468» EMS 5.3.: Bow... 02 8.822 Haesofiaofi 522:3
.3895 R6 «go :3 oz 33%: as;
3 SEE an 52: ED Santana . 03£I>
gm 8:8“ mans—qua:

Source: Calculated from the data

70

6.2.2 Estimation techniques

Using the explanatory variables above, Tobit, Heckman, and Cragg double hurdle models
are estimated using each country’s household survey data. For Zambia, the first and
second supplemental surveys are used (data fi‘om 1999/00 and 2002/03, respectively),

and in Mozambique the 2002 TIA and the 2005 TIA are used.

Each econometric technique adds insight to the question of maize market participation,
but each has its own limitations. It is beyond the scope of this paper to identify the best
econometric technique to fit the question at hand. Therefore, the paper utilizes the
available qualities of all three econometric models to triangulate qualitative and
quantitative effects. No attempt to pinpoint exact magnitudes of explanatory variables is
made, but partial effects are calculated where possible to indicate approximate

magnitudes

None of the coefficients from these three models can be interpreted as partial effects.
Direct comparison of results thus requires post-estimation computation of the
unconditional partial effects. Guidance for calculating UAPEs follows Wooldridge
(2002, p. 543) for the Tobit model and Burke (forthcoming, 2009) for the Cragg double
hurdle model. For the Cragg model, standard errors are calculated and bootstrapped to
calculate significance for the unconditional average partial effects. For the Heckman
model, programming the statistical software to compute average partial effects is

possible, but beyond the scope of this paper. Therefore, the results from the Heckman

71

model can only be relied upon to help indicate direction and significance of results, and

to compare with the output from the other two models.

In the results tables that follow, it is clear that the first stage of both the Heckman and
Cragg models return identical results. This is because the first stage of both techniques
utilizes a Probit model on all observations using all variables. Marginal effects of the
Probit model are utilized for interpretation of the results in the next chapter, and are listed
in the tables. The marginal effects from the Probit model are the best estimation of the
probability that a household enters the maize market given a change in the variable of

interest.

72

7. RESULTS

Key variables expected to influence maize marketing include cassava hectares (hectc) the
kilograms of maize harvested (kg_harvm), price of maize (pricem), kilometers to primary
roads (km_road), and seasonal rainfall (trmm) (Table 10). Additional variables are
included to allow the discussion of the primary variable in a ceteris paribus context. The
following discussion is a detailed description of the results for the cassava hectares and
other key variables, and the following subsection briefly discusses results from some of

the additional variables.

Because several econometric techniques are used on several data sets, this section is
separated by country and by year for convenience of discussion. The data for Zambia
represent two approximately normal rainfall years, in which no wide-spread maize crop
loss or unusual maize price spikes occurred. However, in Mozambique, the data
represent one good rainfall year (2001/02) and one poor rainfall year (2004/05). These
differences in rainfall between years provide further insight to the results presented

below.
To facilitate the discussion of the main results, the tables included in this section only

contain the variables of primary interest to the study. The complete tables containing the

output from all explanatory variables can be found in Appendix 1.

73

7.1 Detailed Discussion of Primary Variables

7.1.1 Zambia: 1999/00 (good rainfall year)

Results from the Zambia 1999/00 survey year for the hectares of cassava variable (hectc)
show a generally positive relationship to kilograms of maize sold. Positive and
statistically significant results are obtained from the Tobit model, the first stages of the
Heckman and Cragg models, the marginal effects from the Probit model and the
unconditional average partial effects (UAPEs) for the Cragg model. In each of the two-
step models, an additional hectare of cassava is positively related to entering the maize
market with marginal effects of approximately 8%. UAPEs from the Cragg and Tobit
models indicate that a household who cultivates another hectare of cassava will sell from
between 60 and 222 kilograms more maize, respectively. Contrary to expectation, the
second stage of the Heckman is negative and significant and the second stage of the

Cragg is positive and insignificant.

Maize price (pricem) and maize kilograms harvested (kg_harvm) are both significant and
positively related to maize sales in most of the model results. As with the hectares of
cassava variable, and contrary to expectation, the kilograms of maize harvested is
insignificant in the second stage of the Heckman model. UAPEs for both variables are
reasonably small, indicating that a small change in price or harvest will have a small
change in maize sales: an extra kilogram harvested yields between .17 and .5 kilograms

sold.

74

The results overwhelmingly confirm that distance to primary road is negatively related to
maize marketing. In all cases, the coefficient on the distance variable (km_road) is
negative; and with the exception of the second stage Heckman, the results are statistically
significant as well. Magnitudes from the UAPEs of Cragg and Tobit indicate that for
each additional kilometer, a household can expect to sell between .6 and .8 of a kilogram

less maize.

The Tobit model and the first stages of the Heckman and Cragg model indicate a positive
and significant relationship between accumulated rainfall (trmm) and increased maize
sales. However, the second stages of both the Heckman and the Cragg models show a
negative result, and the UAPE for the Cragg model, while positive, is not statistically
significant. This variable is an attempt to quantify the observed differences in crop
mixture and sales behavior outlined in earlier sections. It seems that while living in
higher rainfall areas boosted the probability of maize market participation, it does not

necessarily increase the quantity of maize marketed.

7.1.2 Zambia 2002/03 (good rainfall year)

Results from 2002/03 are identical to 1999/00 for hectares of cassava and kilograms of
maize harvested (Table 10). Marginal effects of an additional hectare of cassava increase
the probability of market participation by nearly 10%, and UAPEs indicate an increase in
maize sold of between 51 and 178 kilograms. As in 1999/00, the marginal change in
maize sales following a one kilogram increase in maize production ranges from 0.15 to

0.48 kilograms.

75

Zambia key results

Table 10

1.8!. EEh 2.. Ea...
=85! .ucouuoa veg-SEE: n EC... .30. 32.5.... 0. 28082... n .9335. .09.... 0.3!:
580.... Eben. u 80...... #8993... org-E .o 2550... n Eluslux .2638 .o 83...... n at... ”ulna_.a>

 

 

 

 

68:. 255% noun-bacon 5:5 05.50.? ago» 9820 .3.- nuotm .53.. .623... 1:020:93: a
9.2.2::- fo... «.320 c. not «a.— Eo... .395 E2. 3050 759-! s
.185 2.75.:- goo» 9820 o. o... 3.5:... .288 o
1.3.: 5.2:...- ..o£ ’82.. o. o... 8.3.2.3... .2... m
1.6:. not 2.... 5.5.3: 5 a... 5553... been.» v
.82: on... 2s. 5.5.8: o. o... .8_.8_8. .2... o
2.x. . .89.. .. .59.. Bot...- noozm it... 829.. 38.58%: N
8355.... E 22.0 2558 .82: .30.— Eo... u._:aow_.l_.
.899... C88... .55. .388... 988...... 288...... .23.!
53.3.. 52.8.... 3...... ....=~8... £38.... 52%.... man ............ 22..
.28.... .58... .83. .988... .383. 358...... $9.... .
£385. ..........o...... 2:32- ...ot.....o- :98: £2.89..- 2...... :33. 32:5.
358... £88.... .89.. .938... 8883. 638...... 99......
ha?“ w.u cuumgqo augtmu caughOSAU INSEWOI ac-ENS0.0 N3.0 I-quo Fun—ha
.ano... :88... 8:... efioooo... .8889... .8883. .88....
:38: ......:........ .....x... 58.88... 32.8.... ...m........o... 8.... 58.... £221....
.329. :35... 3.8... .33.... 8:... .83.... .28.
...§.... ...2.8..... 3.9. 5.3... :53..- ...:.~... :3: ......8... no...
32....
.88.... .808... .89... .888... .338... 688...... 5......

.58.... 5.58.... mood- 2.598... .........8.... ............... m8... icon. .55.
.83.... 888... .83. .888... 9.3.8... 39.89... :9... ..
3.8036- :oROOOd- oI-wfifiw- :owmoOSdI mB§.Ou-=0880.0u had- gnomNN.F. filo. Ex

83...... .58.... .98.... .883... 628...... .888... 3......
is... .38.... is...” .888... .85.... .982... 3.... :3..." 58....
888.... .308... .32.. 3288.... .98§.o. 6.58.... 3.8....
$38.... ...28..... 59.2 ...338.oo.o88....- ...3..8o.o 3.... .59.... £22....
:33”. .92.... .28. .38... 8:... 63...... .33.
3.8%.... :35: no? Soon... .58...- room... 8......" 533 so...
was. 2.2:”... .35an 955.5 can. .3... 6.5%.... no.3 .33 oeooo.
- . o n o n u .. sflzfi

 

L

76

The maize price variable for 2002/03 shows a mixed picture. It is positive and significant
as expected for the Tobit model and the first stages of the two-step models, but negative
and significant in the second stages. UAPEs give a mixed picture of the impact of maize
price, with the Cragg UAPE showing no increase in maize sales given a price increase,

and the UAPEs fi'om Tobit showing a half-kilogram increase.

As expected, the kilometers to road variable is negative and significantly related to maize
sales in 2002/03, with the exception of the second stage Heckman output which is
positive and significant. UAPEs from Cragg and Tobit indicate that an extra kilometer of

distance between farmer and main road reduces maize sales by nearly 1 kilogram.

Accumulated rainfall is positively related to maize sales in 2002/03 in all cases apart
from the second stage of the Heckman, and UAPEs indicate between .8 and 3 kilograms

increase for a marginal increase in rainfall.

7.1.3 Summary

In both years, the data overwhelmingly indicate that hectares of cassava are positively
related to maize sales. This result is consistent in each case apart from the second stage
of the Heckman model. Because the two-step models include the market participation
and quantity decisions, it would be beneficial in future work to calculate UAPEs for the
Heckman model for comparison with the Tobit and Cragg. With the information here, it
is unclear whether UAPEs from the Heckman model would be positive or negative,
influenced more by the market participation or quantity decision, but given the strongly

positive results from the other models, it seems likely that the UAPE would be positive.

77

As expected, kilograms of maize harvested and maize price are positively related to
maize sales in most of the model outputs. Although discrepancies in this result exist,
particularly for maize price, this can be explained by the design of the price variable. The
price variable reflects the district median price for each year which does not perfectly

reflect producer prices or the price variability at the household level.

Kilometers to main road for both years are a negative influence on maize prices. This is a
logical result, reflecting the increasing costs of market discovery and other transactions
costs associated with remoteness, and strongly advocates for continued use of GIS as a

tool to create accurate spatial data.

Finally, the accumulated rainfall variable indicates that marginal changes in rainfall do,
ceteris paribus, have a positive relationship with maize sales. It is likely that
accumulated rainfall is not the driving force behind this relationship, but rather it is

reflecting the effects of unobservable variables highly correlated with rainfall.

7.1.4 Mozambique 2001/02 (good rainfall year)

Compared with Zambia, the results for Mozambique paint a mixed picture of the effect of
cassava hectares on maize sales (Table 11). In 2001/02, results indicate that cassava
hectares played little to no role in maize sales, while in 2004/05 the results are mixed.
Among the other key variables, price of maize, kilometers to main road and accumulated

rainfall also return unexpected results in some cases. The substantial rainfall difference

78

between the two years is likely to have influenced the results and is potentially important

to understanding the relationship between maize sales and the key explanatory variables.

Results for hectares of cassava in 2001/02 are not statistically different from zero except
in the second stage of the Cragg model. While this result differs from expectation, and
fi'om the results obtained for similar years in Zambia, it may be a reflection of
substantially different, unanticipated and unmeasured circumstances in Mozambique

compared with Zambia.

Maize production and maize price are both positively related to maize sales for this
dataset, although maize price is not statistically significant in the Cragg UAPE. An extra
kilogram of maize harvested is predicted to yield between .07 and .27 kilograms of
additional maize sold. Holding all things constant, it is clearly logical that larger harvests
of maize will return higher average sales. That price is not strongly related to sales may
again be a reflection of the little variation in the price variable. As in Zambia errors in

farm-gate price measurement led to the use of district median prices for this study.

Unexpectedly, the distance to road variable is positive and significantly related to maize
sales in most of the results for this year. The Tobit model and the first stages of the two-
step models indicate this positive relationship, but encouragingly the second stages and

the UAPE for the Cragg indicate negative, albeit not highly significant, results.

79

Like the distance to road variable, accumulated rainfall is positive and significant for
Tobit and the first stages of the two-step models, but negative or insignificantly positively
related in the other results. Because of the nature of this variable, being a proxy for

unobservable variables related to rainfall, interpretation of this mixed result is difficult.

7.1.5 Mozambique 2004/05 (below average rainfall year)

Compared to the previous year for Mozambique, the 2004/05 year more closely
resembles the results from the two years fi'om Zambia (Table 11). The Tobit model and
the first stage of the two-step models are positive, but the second stages are weakly
negative. While the UAPE fi'om the Tobit model indicates a 72 kilogram increase in
sales with an additional cassava hectare, the resulting UAPE for the Cragg is not
statistically significant, likely resulting fiom the rather small (less than 5%) increase in

probability of market participation found by the marginal effects after the first stage.

As with all previous years, the maize harvest is positive and statistically significant in all
but the second stage of the Heckman model. UAPEs from Tobit and Cragg indicate a
modest increase in maize sales from between .08 and .3 kilograms with an additional

kilogram of harvest.

Contrary to all previous years, the price of maize variable in Mozambique for 2004/05 is
negative and significant in a number of results, and positive and insignificant in the
remainder. This is clearly contrary to expectation, and is again likely to be a
consequence of low variation in the district median price variable used in lieu of accurate

farm gate prices.

80

.386 E55 0.... E2.
=15! inane-n tea-.383... u En... in... basin 0. 28062... n .3035. 6th 3.2:
58!... 839-. u 83.3 .1832... 03...: .0 02.60... u 2.5.0ng .2638 .o 33...... u Bu... 8.333)

 

 

81

 

 

 

.28.. Estate noun-.508. 5.3 252:...- mno... 9820 .3.- ..uotm 3.3.. 09293. 8:020:08: a

5.2.2.0.- .Boh 0.320 c. an... «a: £2... 32¢ Ea... 32!. 859-5 -

38... £85.... .33 9820 c. a... .553... 38.6 o

1.8.: 5.2:...- ....£ .820 c. a... .8882... .2... m

1.8:. not 2.... 5.5.3... 5 no: 5.5:... 9.8.» v

.82.. 8... 2... 5.5.8... c. a... .8383. .2... a

88 . 8...... .. 5...... .33....- asmm 3...... 823.. .23....825 N

8355.! 5 22.0 21.53 .80.: .3...- Eoc 253... F
.882... ”88...... 2.5.8 4:58.... 688...... 458...... .58.

unmcflmof cub—.SOAU alga-w? uuuaggd 2:8 w8.0- 3.82806 NPNV.O nah—.QO SE:
683. £88.... .85.... .888... 2:8... 688...... :8...

ccamwaflf ana8°O°dn 2386?: :60 F86: unwano .38 P006- “vagu- ulbmhén VISIEJ
.538... .588... 88.. .8 5.... 68...... .85... 6.. :.

8 5...... 2.88....- 38 :88..- 8...... .58....- mi..- .85.- 88....
.858... :88... .52.. 6888.... 638...... .8886... .98....

£82.... £88.... :88... ...888... .88....- ...8..8.... 8...... :82. 522......
m 2.89:. .285... .m 8.... .88.... .82.. 68...... 693.

m OVEN.” Ia-§°.O qgmu unlwfi.° a§.0u :QFNN.O DE. Pb Inb.g—. 8.0.:

m 8.58
v. 6888.. .888... .8... c .8588 .88.... 698...... .88....

I“ NFQ—mflmo. cacw§°.° 3‘.Nu «IIMOF8.° I-“§.°i caIflQFOOAU ”8.0 Ic-aqo Egu

m 688... .588... .83. .888... .88.... 688...... 68.8 I

m 558...- 58.3.... .83- 2.1.38... .882...- ..N.:8.. .8... .282. .32 .5.
.m 6.88.... 685.... .n 5.... .25.... .82.. .25.... .38.

a 8.2.3 .858... Nut. .88... mi...- .88.. 89.8 .88 88.5
m .8883 .88.... .92.... 6888.... 688...... .2885... .88....

M :88... £38.... £3... ...888... 52.8....- ...8..8.... Eh... 58.8 522......
h 638...: :85... .6. :8 .58.... .82.. .88.... .58.

u 88...... 88.....- ...$... 85...- on: 85...- 2-3 8.: no...

N... mug .32..me as... an gamma can. .3... 8.7.8.. m8... .38 35....

n n m o o .. n ~ F 806556...

 

 

Like both years for Zambia, the kilometers to main road is found to be negatively related
to maize sales. In all but the second stage Heckman, the results are negative and highly
statistically significant. The UAPEs suggest that for an additional kilometer separating
the household from the main road, a reduction of between .3 and .7 of a kilogram is

expected.

Lastly, the accumulated rainfall variable returns mixed results for this survey year. The
Tobit and first stages of the Heckman and Cragg models are positive and significant,
while the second stages are negative and significant. The Cragg UAPE indicates that the
reduction in quantity outweighs the potential increase in probable market entry, returning

a negative and significant result that is contrary to the UAPE of the Tobit model.

7.1.6 Summary

Unlike the results for Zambia, the Mozambique data do not suggest an overwhelmingly
positive relationship between maize sales and cassava hectares. However, that the good
rainfall year (2001/02) is less responsiveness than the high rainfall year is consistent with
reports from farmers that during times of poor maize harvest, associated with meager
precipitation, cassava is heavily relied upon as a substitute in consumption. Households
with both crops would therefore be in a better position to sell maize, relying on their
cassava stocks and taking advantage of the fact that non-cassava growing households do

not have a substitute.

Price in the first year is positively related to maize sales as expected, and although the

price variable in the poor rainfall year is negatively related to sales, contrary to economic

82

theory, it is possible that the combination of an unusual marketing environment sparked
by wide spread maize crop loss, coupled with the insufficiently accurate price variable is

responsible for the result.

As in Zambia, increased maize production in both years of the Mozambique data is
positively related to sales; higher production indicates that a household is more likely to

satisfy its food requirements.

That kilometers to main road is negatively related to maize sales in one year and
inconclusively related in the other is an unexpected and unexplainable result. It is

perhaps the result of a combination of unobserved variables.

Unlike Zambia, the accumulated rainfall variable is inconclusively related to maize sales
for both years in Mozambique. This may indicate that factors attributable to rainfall in
Zambia are not equally important in Mozambique. Given the different historical,
cultural, and agro-ecological climate in the two countries, perhaps this result should not

be surprising.

7.2 Brief Discussion of Auxiliary Variables

Although the main purpose of the study is to discuss the relationship between cassava and
maize sales, it is helpful to look at the other variables as well (Refer to the Appendix for
the complete set of results). The pattern that emerges most clearly is that the need to
satisfy household consumption is of paramount importance. Variables that are related to

food supply or consumption are generally a significant factor for maize sales.

83

 

 

In the models, variables that mostly seem to positively affect maize sales are total
hectares available to the household, and bicycle or radio ownership. These variables
have the expected, positive sign in most cases. Having more maize and access to land
indicates that a household is more likely to be able to satisfy its food requirements and
owning a radio or bicycle indicates a household has better access to market information

and can transport its own produce.

The variables most often negatively related to maize sales are the number of household
members and the age of the household head. This former indicates that additional
household members require higher quantities of food to be retained for consumption,
thereby leaving less for sale. The latter suggests that older household heads may not be

as able to search or negotiate for maize sales opportunities.

84

8. CONCLUSIONS

Involving smallholder producers in the maize market in Zambia and Mozambique has the
potential to increase the food supply and stimulate economic growth through increased
rural incomes and linkages to other sectors. Previous experience and research indicates
that for households to produce food for sale in the market, a number of requirements must
be met. First and foremost, household food security must be attained. Empirical studies
that explore the determinants to market participation indirectly address food security as
an important factor of participation. The major contribution of this study is that it
explicitly includes the effect of cassava hectares as a substitute food staple on the
estimates of household maize market participation, controlling for socioeconomic,

demographic, and market access effects.

By evaluating results fi'om the Tobit, Heckman, and Cragg models together, a general
picture of the importance of cassava stocks to maize sales emerges in Zambia and
Mozambique. The results indicate that in Zambia, cassava hectares are positively related
to market participation and level of maize sales. For both years, UAPEs fi'om the Tobit
and Cragg models return positive results between 50 and 220 kilograms of maize sold per
hectare of cassava. This increase is largely driven by the increase in market entry that
cassava stocks make possible. Given that both of the datasets for Zambia represent years
of good rainfall, and normal maize price fluctuations, it is likely that cassava stocks
increase maize sales in all years. In poor maize years, it is thought that households would

rely to a greater degree on cassava than in good years.

85

In Mozambique, results indicate that cassava hectares have little impact on maize sales in
good rainfall years and greater impact during years of little rain. In the good rainfall
year, only one result was positive and significant for the hectares of cassava variable,
while in the below normal rainfall year, several results were positive and significant. As
in Zambia, the bulk of the positive values relate to the probability of entering the maize
market as a seller. However, because the UAPE for the Cragg model is not statistically

significant, a positive relationship is not strongly confirmed.

The Mozambique results are consistent with household planting regimes that stagger
cassava planting to buffer maize shortfalls from one year to the next. Because cassava is

planted two years before it is consumed, and because it is resistant to poor rainfall,

households may depend on the crop more during bad maize seasonsl 1.

8.1 F inal Comments

This study demonstrates the importance of cassava hectares to maize market participation
in Zambia. This information should help policymakers to prioritize agricultural

spending, perhaps with renewed emphasis on diversifying smallholder staple crops. In
Zambia, while cassava has long been recognized as a staple food crop, the potential
consumption smoothing benefit of cassava production, leading to increased probability of
maize sales, has largely been ignored. Likewise, compared to areas along the line of rail,

large areas of Northern Zambia, where both maize and cassava are grewn, remain

 

'1 Indeed, cassava breeders indicate that farmers may be reluctant to give up long-
duration varieties, which they can harvest when they like over a multiple year period,
because the long duration of maturation provides an extra measure of risk aversion for
farmers (Steve Haggblade, pers com).

86

geographically marginalized by poor infi'astructure. Perhaps due to its perception as a

subsistence farming area, policy has not focused on this productive area of the country.

Given the results from Mozambique, policymakers should note that cassava’s
relationship to maize market participation increases in times of national maize deficit.
Although the relationship in Mozambique is not as strong as in Zambia, increased focus
on the interaction between staple foods seems warranted for both countries. Given what
appears to be cassava’s buffering role in household consumption, continued efforts to

promote disease and insect-resistant cassava varieties is justified.

Though this study provides a tantalizing glimpse into the interrelationship between
cassava and maize sales, a number of questions remain that call for firrther analysis.

First, while this study attempted to utilize available econometric techniques to estimate
its findings, one shortcoming is its inability to compare unconditional average partial
effects between all three models. Future econometric work should focus on evaluating
UAPEs for the Heckman model to compare with UAPEs from the Cragg and Tobit
models. Second, the results from Mozambique showing differentiations between good
and bad maize harvest years suggest that cassava’s importance to maize sales is variable
depending on the adequacy of maize harvest. Further studies of this kind should compare
producer prices in good and bad maize production years to further test the extent to which

the ability to substitute maize for cassava increases in importance as maize prices rise.

87

 

APPENDIX

88

 

.58... o8... 8...... 88.. 8...... .88.

 

Table 12: Zambia 1999/00 complete results

.93.. :9... in... 1.9.... 28.... £88.
.. 8...... 2.5.. .53.. .88.. .35.. 3....
.88.. IE. .28.. 88.. .28.. :3:
.88... EB. .86.. .283 .858 can.
.155... .83 LB... .58... LB... £88
.35... an... Ex... 83 5...... 8....
.188... :3... :93 8.8... .13... L3.
.68... 8:. £8.» 88... 388... .3...
L88... 8.: LES... £28.. 1.8.88 .259.
985... 38. .88.. .83 8.8.. .53.
88.. 1.8.. 88.. 88.. 88.. I8
.58... .88. .58.. 8.83 .8889 38...
£8.89 88- :8... La... rfia... 1.31
.88... .888... 6.8:. 8.8... 6.3.... 8.8.8...
138888... £8.88... 3.3.8.8.» 88+ 1.3.888... £8.88...
.38... .88... .3... 5.98.. @883 @288... 58...
2:8... 228.. 38...... 538... 9.8.8.9 1338.. 8.... 3.9:.
.88... .58... «B... .88.... .8289 888.. can...
:55 .88.. is}. .88... .88.. 8.8... :3 £8...“
.88... $88... 8.... .88.. 88.. .88... 38...
in..." .98.. :82. 1.3.8.. .88.. :98. 8.... 1.3.
.89... 688... can. 3.88.. 98.89 .8889 99...
EB... :38... 8.8. 3.88.. L... .8... 2.8.8.. 8.... :2.
.88... 688... 98... 988.... .88... £88... .82..
18.8... 2.888... Ln..- .8.qu 5.8... «.858... 8... LE-
.Efi. .35... «N... 6.8.. 6:... .98... can.
:88 -88.. 88¢ .18... L8... 28... 8.38 in...
g .5... .8... .3... can.
188.. Lawn. 3:8. .588 .89.

mt: lynx"! Eula id's In?! 858... PI. .38

a . .. n v n ~ ..

 

 

89

6......

.39..

.83....

 

6......

69°...

.83.

 

 

 

 

 

188.. .8. ...E... 8.... L..... ......m. as
.8. a8... .82.... 8.... .85.... 889
.88.... .18... 8.. 8... ...E... ...-...... a!
.8... 8...... 8...... 8... 8...... ......u
8...... 8.... 8.. 88...- ...88 8.... 8......
.58... . ...-w... 8...... 8.... 8.8... .83
8...... 1878:... m..... 8:... a...” .25.
.38.... ...... .... .88... .88... 8...... n3. ..
.138..- 8... 5.88..- 8.8.... .88... 8... ...-
.maa... .38. 88... .3... .88... 8a..
888... .8... n8. .8...- Ba..- .58..
.88... 6.. B. .58... 6.5... .88... .88..
1R8..- 8...- :8..- 5.... ...-88..- 1.83 .5...
...8... 38.88... 88... 88.8. 938... .888...
s £88.88... 5888... £8.88.- 8. 1.8888... .88... 8...:
m 8.8. .8... 6.3. .88... .888... .888... 8...
m 338.... than... ...-...... ...-8.. 8 £88.... 8.... .8... Er...
.... «.8... .88.... ..mu. «8... 6.88... 688.... ...-8..
k 8.8..- 8 38...... ....E8.. .88... 2.58.... to... :8... :88
m 3...... 8...... Es. .8. 8.3... $8.... and
w 8... ...-:8... ......8- rL.......- 2.8:... £85.98- ...88- ...a
m 86.... .68.... ..R... .88... c8... .88.... 68..
W 12...... 3.38.. 1.8.... ...-9.8.. ......8... 88.9.... ...-8.. E...
m. 58.... .88. an... .88... .8 .88.... .8...
.m ...-:8...- ...8.8... .8...- 8- 8.8.... 8.8.9.8.. 88 82:5.
.m 8.... =8... ......8. .88... as... .85... .83..
h 1.8.... 1.5.8.. ...... 2.8.. ins... 1.8.8. it...” as:
.. 8.... 8.... .88. 8.... we...
...-... tum..- ...-...... IE... ...-.....- ......N. 8
...m. ml: 8...! EM... 8:85 Ema-... ...-...8: ...—5 I... 8
n a rm ...7 m .. m T . 8

 

 

91

Table 13: (cont’d)

again.
.2! igufifiufiiiefliixuleas SEE

iggufifigitéugfiitiuiii.

EigigfifluE-oitfifi 1585435595 m

Ezuiagcseiwecaeicfigi .

391.313.38.0s8i 2.1.18...- a

Egipagsfii.iisum m

8:32.553. 33....qu ..

gaugiigfligfiolut n

5.11%.... ...-aging: n

 

13:89:23.! .
“a I Hi i .51 'l
£5.23: :8 .88 ...8 n8 ...-8 .6338
38. .35
...mN i8... :51
as... .
Inn-n. it!
g... 0.3. E... Em... c8... an...
5...... 35- RN..- 8.... ..N... «8.- 3
ENE... 9.9. 8.... a... 3.... .18...
308.. .188 2...... 8.... 98.. BB 9!:

92

Table 14: Mozambique 2001/02 complete results

 

 

3.... «.8... 58.8.... 6.882.. c.3888. .38...
g... .9...» 5888... 88.82.. 338.. 88... .8!
...8... 8...... $888.. 9.58.3. 9888:... .85...
E... SE... 9.58... 8.88... 8.88:... .99.... 8E!
68... 3.8... 92.3... 9888:... 8.888. QB...
$588.... 85.9 g... 83:... £8888- 1.3.... 5......
.888... .38... .28... .23... .95... 9.8....
3.88... ......N. .58.... 6...... 3.8a... £89 38.
.88... .89. .85.... 088... .858. ...... ..
......8... 8.... :3... in... than... 1.8.... ...-
.83... #8.. .88... 6K... 68:... .83.
.88.... £89 ...»... 389 $8.. .5... 582
.38... .83 g... «9...... .888. .59
:88... 8...... .288... 8:... loan... in... ...-E
.83. 9.8.8... 9.1.8.. 5.8m... .382. 6.8.3...
£38.88... tax—...? ...8835.? Ed .8823... £53... «.52....
6888. ...—.8... 6...... £38.. .583. 588...: .88...
1.3.... $28.... .135 .1888... 8.8.3. 1.8.8.. .5... 18.... Ed.
.....82... .85... 6...... .28.... 6...... 6.8... an”...
8.3.... .28.... ...... .88.. 9...... .93.. 8...» ....NB :3...
€88... .....8... 6...“. .88... .85... .858. ...... ..
28.....- .158... -89. 3.88... 35.. £583 .8. .- .....Bv ....
63.8... .38... .3... .858... .85.... 5.8.... .88....
88...... £38.... 35. L98... .58... :38... 3... IE... . 5:...
SB... ...—.8... .8... 68...... 8.8... 688.... fine
.38..- 33...: .98.... L388 .883. 1.9.8... .8... $8.. 82......
8.8.... ..85... ... .... .58.. 68... .55... .52.
88;. 8.8... :8... 8...... as... 8...... nu... mm... an...
....N... .5... .92. .52.. as...
.23.... ......m..- 5...... 2.8.- La... .5150
ES .35.! is? a... Emu In?! 8...]. ES 8.... 3.8...
a . o .. w n a p g

 

93

Table 14: (cont’d)

iv... .89.. 1.59.. ....

Ull- EEE. E

It! :33 nit-Bun EE..!EE Sub!!!" 2:5. .BEaUE
ggufigiflst§u§£§.§a§§u 5231!,
«big nan-ins 52215.: 32.30 5.3.35 rigging a

Essa-053355; .Bnficnufi En! »
loans-5... ...E .35 c. ...... Evussu o
Escapism... 5923553. ”5.3253 Em m
32332553.. 2.5.52.3 ...

18...}... gsfigsnfiififi ...... n
Ehaiuflfiuiaanmiiiisufi N
88.325553: P

 

int-{£553.86 as. g g g 89 2.1630
as: ”9.2.
182 1.3:. Bus
fits
L8". .3!
3853 go End :8: Ewe was
Eu :3 83 :3. an... 95. .25.
:98... 6.8: 9.3 :89 6:9 A85
928.. :8: ma... 53. Sud :28 .3.
533. as; find 63: fine 69:.
..EB... ham. 8...? E8... 88. has”. as...
359 68... £93 EN... sane Eda
...Nqo 3e. .28... 88a. .28.. 5.8 a...
:88 a 38.9 989.9 6282. age 38.3
9333 5.8.... an E... 935.: BB: :5.

94

59 $5.8 g. $286.3 Ed. 6389

 

$.38... ..88... 2888..- ES. 2888..- 1.9.8. .3!
.82.. $8... .....88... .388... .....88... 5......
.38....- mu... 8:3...- ...88..- 8.88..- .....a. «as:
g... ...... 9888... 888... £888... 02....
...-.589 .8..- g..- 8.8.... III... 8...... .35...
6.8.... ...-E. SE... .88... .85... ...... ..
.85....- im... 58... .s... ... 58..- 8... ...—....
.58.... «...... 30.8... 998... .38... .8..- ..
58.8..- :88. . 8!... 5.68..- ...-x... .. 8..
...—.6... an... ...-58... .5... $8... 6....»
.35.. L8. .8... a 88..- En ...-.... 6......
3.8... 5.8. «88... .88... «88... 6......
s ...“...8... a... £88..- .83.. 18!..- 1.5.... .5...
m 3.3 5.88... .83.... 8.8... .83.. .n. 688......»
m ...-8533.... .1858... $88... ...-88.8.... Luna—.8 ..- ....Euaao. 8.22:9.
m 68.8.. £82. ..8... .588... 6.88... .838... ...-.8...
m. ...-6.... L388. ...-.3. fig. .. 888..- ...-398.983.. :88... 522......
m was... cgne ad... .85... .88... an... ... .8. ..
w .28... 23....- n8... :88..- 8...... 188.993..- .8. .- 53....
M g. 63...... ...-NE. ...-.8... A. :8... 5.8... 6.... ..
W :5...- 283... ...-8.... 858..- 88.... 8588-88.. 8... ....
m 6.8.... ...—...... 3nd ... ...8... .888... ......8... ...-....
m :33.- i.....8.. LN..- E... ...-8.8..- E35... ...-...... .5...
.m. 35.... 93...... 98... $88... a ..8... 83... .53
.m £88..- :83... 38...- 58.8..- $5.... ...8.8..-......- 1...... .- 82......
n .28.... 6......» 6...... 6.8... 5.... .88... an...
m 383... ...-35.. mm...- ifi. .. a... law: as... in... an...
.. in... 6...... 5.... 9.... 98 ..
H ...-mm...- .-...~n.- £8... :8...- Lfiv .525
m mi: 32...! d5... mu 9...? E8... 8...}. mt: ...... as...
n a m o ... v m a .. gnu...

 

 

 

 

95

Table 15: (cont’d)

...-1551......

in... Iiiaifiufisua.gsaui8:uie|§ 8881..
igniuufiiilitganfilfi (fixings;
gala-13:51.21... mt..8.c!..uam 356515.158: m
£§.u§os&ul§.. «88.5.8813 .
1.298.223.36533I538m 8
39.151830305538831... m
13:335.: 8.55.863 v
lanaififiizssfigoluc n
82......qu 1.588.515.5695 u
32.8.5.3 .

 

81.. 8.9.. .- .59....

gsgug 8... 88 88 88 86 23-630

. dd... SE
1...... La ......

.5...

1......- .35

3.8... .88. 8.... ..8... a... 3....
REE. :6...- Ed 8...- 8a.... 8....- i...

.22.... 8... an... as... an... ...8 ..
98.8..- 198 8.... 82. 8.... as... 91..

.85... $8. .85.... 9...... .88... .3...
lap... 1...... :8... 88-18.... 3.8. a...

3.... ....a. 3.58... .938... .258... gm...
it.

go- ofid 9 Ed. Ea EEO- Ed-

96

REFERENCES

Barrett, CB. 2007. “Smallholder Market Participation: Concepts and Evidence from
Eastern and Southern Africa.” For FAO workshop on Staple Food Trade and
Market Policy Options for Promoting Development in Eastern and Southern
Afiica, Rome, March, 2007.

Barrett, CB. 2002. “Food Security and Food Assistance Programs.” Handbook of
Agricultural Economics. 2(2): 2103-2190.

Bellamere, M.F., Barrett, CB. 2006. “An Ordered Tobit Model of Market Participation:
Evidence from Kenya and Ethiopia.” American Journal of Agricultural
Economics. 88(2): 324-337.

Boughton, D., Mather, D., Barrett C.B., Benfica, R., Abdula, D., Tschirley, D., Cunguara,
B., Fall 2007. “Market Participation by Rural households in a Low-Income
Country: An Asset-Based Approach Applied to Mozambique.” Faith and
Economics 50, 64—101.

Burke, W. J. 2008. "Estimating and Interpreting Cragg's Tobit Alternative Using Stata."
Staff Paper (number forthcoming) Michigan State University, Department of
Agriculture, Food, and Resource Economics. East Lansing, Michigan.

Center for International Earth Science Information Network (CIESIN), Columbia
University; and Centro Intemacional de Agricultura Tropical (CIAT). 2005.
Gridded Population of the World Version 3 (GPWv3): Population Grids.
Palisades, NY: Socioeconomic Data and Applications Center (SEDAC),
Columbia University. Available at http://sedac.ciesin.columbia.edu/gpw.
(Accessed January 3, 2007).

Collinson, M. Farm Management in Peasant Agriculture. Boulder, CO. Westview
Press.

Cragg, J .G. 1971. “Statistical Models for Limited Dependent Variables with
Applications to the Demand for Durable Goods.” Econometrica, 39(5): 829-844.

De J anvry, A., Fafchamps, M., Sadoulet, E. 1991. “Peasant Household Behavior with
Missing Markets: Some Paradoxes Explained.” The Economic Journal, 101 (409),
1400—1417.

Goetz, SJ. 1992. “A Selectivity Model of Household Food Marketing Behavior in Sub-
Saharan Africa.” American Journal of Agricultural Economics 74(2): 444—452.

Gujarati, D.N. 2003. Basic Econometrics. New Delhi, India. Tata McGraw-Hill
Publishing Company Ltd. Fourth Edition.

97

Fafchamps, M. 1992. “Cash Crop Production, Food Price Volatility, and Rural Market
Integration in the Third World.” American Journal of Agricultural Economics
74(1): 90 — 99.

FAQ. (2006). “Food and Agriculture Indicators. Country: Zambia.”
http://www.fao.org/ES/ess/compendium_2006/pdf/ZAM_ESS_E.pdf (accessed
October 20, 2008)

FAOSTAT. 2008. Food and Agriculture Organization of the United Nations.
http://faostat.fao.org/site/609/default.aspx#ancor (accessed October 1, 2008)

Feder, 0., Just, R.E., Zilberman, D. 2007. “Adoption of Agricultural Innovations in
Developing Countries: A Survey.” Economic Development and Cultural Change
33(2): 255-298.

Haggblade, S.J, Nyembe, M. 2007. “Commercial Dynamics in Zambia’s Cassava Value
Chain.” Report for Cassava Transformation in Southern Africa (CATISA).

Haggblade, S., Nielson, H., Govereh, J ., Dorosh, D., 2008. “Potential Consequences of
Intra-Regional Trade in Short-Term Food Security Crises in Southeastern Afiica.”
Report prepared by MSU for the World Bank.

Heckman, J .J . 1979. “Sample Selection Bias as a Specification Error.” Econometrica,
47(1): 153-161.

Hijmans, R.J., S.E. Cameron, J .L. Parra, P.G. Jones and A. Jarvis, 2005. Very High
Resolution Interpolated Climate Surfaces for Global Land Areas. International
Journal of Climatology 25: 1965-1978.

Jain, S. 2007. “An Empirical Economic Assessment of Impacts of Climate Change on
Agriculture in Zambia.” The World Bank Development Research Group
Sustainable Rural and Urban Development Team. Policy Research Working
Paper 4291.

Jayne, T.S., Mukumbu, M., Chisvo, M., Tschirley, D., Zulu, B., Weber, M.T., Johansson,
R., Santos, P., Soroko, D. 1999. “Success and Challenges of Food Market
Reform: Experiences from Kenya, Mozambique, Zambia and Zimbabwe.” MSU
International Working Paper No. 72, 1999.

Jayne, T.S., Govereh, J ., Chilonda, P., Mason, N., Chapoto, A., Haantuba, H. 2007.
“Trends in Agricultural and Rural Development Indicators in Zambia.” Working
Paper No. 24, Food Security Research Project, Lusaka, Zambia.

Kokwe, GM. 1997. Maize, Markets &. Livlihoods, Helsinki, Finland. Institute of
Development Studies, University of Helsinki.

98

Megill, DJ. 2004. “Recommendations on the Sample Design for Post-Harvest Surveys
in Zambia Based on the 2000 Census.” Working Paper No. 11, Food Security
Research Project, Lusaka, Zambia.

Nweke, F.I., Dunstan S.C.S., Lynam, J .K. 2002. The Cassava Transformation: Africa’s
Best Kept Secret. East Lansing, MI: Michigan State University Press.

Tobin, J. 1958. “Estimation of Relationships for Limited Dependent Variables.”
Econometrica, 46:1 (24-36).

Tschirley, D., Abdula, D., Weber, M.T. 2006. “Toward Improved Maize Marketing and
Trade Policies to Promote Household Food Security in Central and Southern
Mozambique.” Research Report No. 60E, Ministry of Agriculture Directorate of
Economics, Republic of Mozambique.

Rose, D., Strasberg, P., J eje, J .J ., Tschirley, D. 1999. “Household Food Consumption in
Mozambique: A Case Study in Three Northern Districts.” Republic of
Mozambique Directorate of Economics Research Paper Series. Report No. 33.

Tschirley, D., Nijoff, J .J ., Arlindo, P., Mwinga, B., Weber M.T., Jayne, TS 2006.
“Anticipating and Responding to Drought Emergencies in Southern Africa:
Lessons fiom the 2002-2003 Experience.” MSU International Development
Working Paper No. 90.

Walker, T., Pitoro, R., Tomo, A., C. Salencia, Mahanzule, R., Donovan, C., Mazuze, F.
2006. “Priority Setting-for Public-Sector Agricultural Research in Mozambique
with the National Agricultural Survey Data.” Research report No 3e, Institute of
Agricultural Research of Mozambique.

Whiteside, M. 2003. “Enhancing the Role of Informal Maize Imports in Malawi Food
Security.” Consultancy report for DFID.

Wodajo, T.B. 2008. “A Double-Hurdle Model of Computer and Internet Use in
American Households.” Department of Economics, Western Michigan
University.

Wooldridge, J. 2006. Introductory Econometrics: A Modern Approach. Mason, OH:
Thomson South-Westem. Third Edition.

Wood, A.P., Kean, S.A., Milimo, J .T., Warren, D.M., eds. 1990. The Dynamics of

Agricultural Policy and Reform in Zambia, Ames, IA: Iowa State University
Press.

99

Zulu, B., Nijoff, J .J ., Jayne, T.S., Negassa, A. 2000. “Is the Glass Half-Empty or Half-
Full? An Analysis of Agricultural Production Trends in Zambia.” Working
Paper No. 3, Food Security Research Project, Lusaka, Zambia.

Zulu, B., Jayne, T.S., Beaver, M. 2007. “Smallholder Household Maize Production and
Marketing Behavior in Zambia and its Implications for Policy.” Food Security
Research Project Working Paper No. 22.

100

   

will llllll'lllllllll lllllllllllllllWill/ill!“
3 1293 03062 l826